Ministry of Higher Education & Scientific Research University of Technology Chemical Engineering Department Comparative Study of Temperature Control in a Heat Exchanger Process A Thesis Submitted to the Department of Chemical Engineering of the University of Technology in Partial Fulfillment of the Requirements for the Degree of Master of Science in Chemical Engineering / Unit Operation By Afraa Hilal Kamel Al-Tae (B.Sc. in Chemical Engineering 2005) Supervised by Prof. Dr. Safa A. Al-Naimi March 2011 SUPERVISOR CERTIFICATION I certify that this thesis entitled "Comparative Study of Temperature Control in a Heat Exchanger Process" presented by Afraa Hilal Kamel Al-Tae was prepared under my supervision in partial fulfillment of the requirements for the degree of Master of Science in Chemical Engineering at the Chemical Engineering Department, University of Technology. Signature: Prof. Dr. Safa A. Al-Naimi (Supervisor) Date: / / 2011 In view of the available recommendations I forward this thesis for debate by the Examination Committee. Signature: Asst. Prof. Dr. Mohammed I. Mohammed Head of Post Graduate Committee Department of Chemical Engineering Date: / / 2011 CERTIFICATION This is "Comparative to certify Study of that I have Temperature read the Control thesis in a titled Heat Exchanger Process" and corrected any grammatical mistakes I found. The thesis is therefore qualified for debate. Signature: Prof. Dr. Mumtaz A. Zablouk University of Technology Date: / / 2011 CERTIFICATE We certify that we have read this thesis entitled "Comparative Study of Temperature Control in a Heat Exchanger Process" by Afraa Hilal Kamel Al-Tae and as an Examining Committee examined the student in its contents and that in our opinion it meets the standard of a thesis for the degree of Master of Science in Chemical Engineering. Signature: Prof. Dr. Safa A. Al-Naimi (Supervisor) Date: / / 2011 Signature: Signature: Assist. Prof. Dr. Hassan W. Hilou Dr. Zaidoon M. Shakoor (Member) Date: / (Member) / 2011 Date: / / 2011 Signature: Assist. Prof. Dr. Kutaeba J. Al-Khishali (Chairman) Date: / / 2011 Approved for the University of Technology Signature: Prof. Dr. Mumtaz A. Zablouk Head of Chemical Engineering Department Date: / / 2011 Acknowledgments First of all, praise is to Allah for every thing. Without his great assistance the work wouldn't have been finished. I would like to express my sincere appreciation and thanks to my supervisor Prof. Dr. Safa A. Al-Naimi, for his constant guidance and valuable comments, without which, this thesis would not have been successfully completed. My grateful thanks to Prof. Dr. Mumtaz A. Zablouk, the Chairman of the Department of Chemical Engineering at the University of Technology for the provision of research facilities. My deep thanks go to Assist. Prof. Dr. Mohammed I. Mohammed, the head of post graduate committee for all the help and encouragement, also I wish to express my sincere gratitude to Dr. Orooba N. Abdullah for her support and helpful advice. Special thanks to Assist. Prof. Dr. Amer Al-Dabagh, Mr. Basheer Ahmed, Mr. Alaa Hussain and Mr. Khalid Mansoor for their help and support. Also I would like to convey my sincere appreciation to all staff of Chemical Engineering Department in the University of Technology and the workshops unit especially the welding workshop. Finally, to all that helped me in one way or another, I wish to express my thanks. Afraa Abstract In this work the dynamic behavior of a plate heat exchanger (PHE) (single pass counter current consists of 24 plates) both experimentally and theoretically and the control of the system were studied. Different control strategies; conventional feedback control, classical fuzzy logic control, artificial neural network (NARMA-L2) control and PID fuzzy control were used to control the outlet cold water temperature. A step change is carried in the hot water flow rate which is considered as a manipulated variable. The experimental heat transfer measurements of the PHE show that the overall heat transfer coefficient (U) is related to the hot water flow rate (m h ) by a correlation having the form: U = 11045 m 0.7158 h In this work the PHE model is found dynamically as a first order lead and second order overdamped lag while the experimental PHE represented dynamically as a first order with negligible dead time value. A comparison between the experimental and the theoretical model is carried out and good agreement is obtained. The response of 24 plates when justifying the design fitness of the PHE, proved that the design is accurate and there are no losses in the energy input. The performance criteria used for different control modes are the integral square error (ISE) and integral time-weighted absolute error (ITAE) where the ITAE gave better performance. As well as the parameters of the step performance of the system such as overshoot value, settling time and rise time are used to evaluate the performance of different control strategies. The tuning of control parameters were determined for PI and PID controllers by two different methods; Ziegler-Nichols (Bode diagram), and Cohen-Coon (process reaction curve) to find the best values of proportional gain (K c ), integral time (τ I ) and derivative time (τ D ). Accurate results have been obtained using artificial neural network over PI, PID and classical fuzzy logic controller. The PID fuzzy controller gave better control results of temperature rather than PI, PID, classical fuzzy logic and artificial neural network controller because PID fuzzy controller combines the advantages of a fuzzy logic controller and a PID controller. The best value of settling time and overshoot were found of 0.432 and 1.0 respectively which represent the PID fuzzy controller, while the best rise time found of 0.077 which represent the PID controller. The lower value of ITAE of 0.0031 is obtained which represent the PID fuzzy controller and to certain the best strategy of control among the others. MATLAB program version 7.10 was used as a tool of simulation for all the studies mentioned in this work. Contents Contents I List of Abbreviations V Nomenclature VI Greek Symbols VII Symbol VII List of Tables VIII List of Figures X Chapter One: Introduction 1.1 Introduction 1 1.2 Control of Heat Exchangers 4 1.3 Aim of the Work 6 Chapter Two: Literature Survey 2.1 Introduction 8 2.2 Dynamic Modeling of Heat Exchanger 8 2.3 Control of Heat Exchangers 11 2.3.1 Conventional PI and PID Control 12 2.3.2 Computational Intelligence Techniques 13 2.3.2.1 Fuzzy Logic Control 13 2.3.2.2 Artificial Neural Network Control 15 Chapter Three: Experimental Work 3.1 Introduction 21 3.2 Description of the Experimental Rig 21 3.2.1 Plate Heat Exchanger 21 3.2.2 Sump Tank for Hot Water 22 3.2.3 Cooling Tower 26 3.2.4 Temperature Measurement 27 3.2.5 Water Flow rate Measurement 29 3.3 Description of the Computer Control System 29 3.4 Experimental Procedure 33 3.4.1 Steady - State Data 33 3.4.2 Dynamic Response Data 34 Chapter Four: Modeling and Theoretical Analysis 4.1 Introduction 36 4.2 Model Assumptions 37 4.3 Energy Balance 38 4.3.1 Energy Balance around Cold Plate 38 4.3.2 Energy Balance around Hot Plate 39 4.4 Control Strategies 41 4.4.1 Conventional Feedback Control 41 4.4.2 Controller Tuning 44 4.4.3 Fuzzy Logic Control 45 4.4.3.1 Introduction of Fuzzy Logic 45 4.4.3.2 Linguistic Variables 46 4.4.3.3 Fuzzy Logic Controller 47 4.4.3.3.1 Design of Fuzzy Logic Controller 48 4.4.4 Artificial Neural Network Control 51 4.4.4.1 Introduction of Artificial Neural Network 51 4.4.4.2 Biological Artificial Neural Network 51 4.4.4.3 Mathematical Model of a Neuron 52 4.4.4.4 Architecture of Artificial Neural Network 54 4.4.4.5 4.4.4.6 4.4.4.6.1 Back Propagation (BP) Algorithm Artificial Neural Network Artificial Neural Network Controller Identification and Controller Stages of the NARMA-L2 model 55 58 59 Chapter Five: Results and Discussion 5.1 Introduction 64 5.2 Open Loop System 64 5.2.1 Steady State Results 64 5.2.2 Dynamic Behavior 66 5.2.3 Justifying the Design Fitness of the PHE 68 5.3 Closed Loop System 70 5.3.1 Conventional Feedback Control 70 5.3.1.1 Control Behavior 72 5.3.2 Fuzzy Logic Controller 74 5.3.3 5.3.4 5.3.5 Artificial Neural Network NARMA-L2 Controller PID Fuzzy Controller Comparison Among PID, Artificial Neural Network and PID Fuzzy Controllers 78 82 86 Chapter Six: Conclusions and Future Work 6.1 Conclusions 89 6.2 Future Work 90 References Appendices Appendix A: Calibration Curves of Thermocouples A.1 Appendix B: System and Operating Conditions B.1 Appendix C: Controller Tuning Methods C.1 C.1 Cohen-Coon Method C.1 C.2 Ziegler-Nichols Method C.2 Appendix D: MATLAB Program D.1 D.1 D.1 Introduction D.2 Open Loop Programs D.3 D.3 Close Loop Programs D.4 D.3.1 Ziegler-Nichols Method D.4 D.3.2 Cohen-Coon Method D.7 Appendix E: Experimental Data of Dynamic E.1 Behavior Appendix F: Calculation of Overall Heat Transfer Coefficient (U) V F.1 List of Abbreviations Symbol AC AI ANN AO APV BP CI Definition Alternating Current Analog Input Artificial Neural Network Analog Output Aluminum Plant and Vessel Back-Propagation Computational Intelligence CPT Crude Preheat Train DAQ de DI DO e Er FL GM G PRC GRNN HE ISE ITAE MLBP MLP MSE N NARMA-L2 NARX NB NN NNPC NS P PB PHE PI PID PS RMSE SISO Data Acquisition Board Change of Error Digital Input Digital Output Error Relative Error Fuzzy Logic Gain Margin Process Reaction Curve Transfer Function General Regression Neural Network Heat Exchanger Integral Square Error Integral Time-weighted Absolute Error Multi-Layer Back Propagation Multi Layer Perceptron Mean Square Error Negative Nonlinear Auto Regressive-Moving Average Nonlinear Auto Regressive with eXogenous Negative Big Neural Network Neural Network Predictive Control Negative Small Positive Positive Big Plate Heat Exchanger Proportional-Integral Proportional-Integral-Derivative Positive Small Root Mean Square Error Single Input-Single Output V Symbol V AC V DC Z Definition Alternating Current Voltage Direct Current Voltage Zero Nomenclature Symbol A CP C Pc C Ph G Gc Gm Gp Gv h K Kc KD KI Ku mc Mc mh Mh pu s S t T ci T co td T hi T ho u U Definition Area of heat transfer Heat capacity Cold heat capacity Hot heat capacity Transfer function Transfer function of controller Transfer function of measurment Transfer function of process Transfer function of control valve Heat transfer coefficient Steady-state gain of the process reaction curve method Proportional gain Derivative gain Integral gain Ultimate gain Cold water flow rate Cold water mass Hot water flow rate Hot water mass Ultimate period of sustained cycling Laplacian variable Slop of the tangent at the point of inflection of the process reaction curve method Time Inlet cold water temperature Outlet cold water temperature Time delay Inlet hot water temperature Outlet hot water temperature Control Action Overall heat transfer coefficient Units m2 J/kg.oC J/kg.oC J/kg.oC − − − − − w/m2.oC o C Volt/oC Volt/oC Volt/oC decibels Kg/sec Kg Kg/sec Kg sec/cycle − − sec o C o C sec o C o C − w/m2.oC Greek Symbols Symbol µ ΔT lm τ τa τc τD τh τI τp ψ ω Definition Membership function Logarithmic mean temperature difference Time constant of the process reaction curve method Lead time constant Cold time constant Derivative time constant Hot time constant Integral time constant Lag time constant Damping coefficient Crossover frequency Symbol Symbol ' ̅ ¯ ° Definition Unsteady state Deviation Steady state Units − o C sec sec sec sec sec sec sec − rad/sec List of Tables Table Table (2.1) represents the articles of controlling the HE. Page 19 Table (3.1) Plate heat exchanger specifications 22 Table (3.2) Description of the experimental rig 25 Table (3.3) Standardized detail of the K-type thermocouple Table (4.1) IF-THEN rule base for fuzzy logic control 28 50 Table (5.1) The relative error (Er) and mean square error (MSE) between experimental and theoretical 67 (T co ) response Table (5.2) Control parameters of PI control 70 Table (5.3) Control parameters of PID control 71 Table (5.4) Comparison of different parameters of PI and PID controllers Table (5.5) IF-THEN rule base for classical FL control Table (5.6) Comparison between the performance of fuzzy logic controller and PID controller Table (5.7) Different performance indices and different parameters of ANN NARMA-L2 controller Table (5.8) The rule base of PID fuzzy controller Table (5.9) Different performance indices and different parameters in PID fuzzy controller Table (5.10) Comparison of different performance indices and different parameters in controllers 71 76 77 82 84 85 87 Table (B.1) System and operating conditions B.1 Table (D.1) Summary functions in MATLAB program D.2 Table Page Table (E.1) ( ∆ T lm ) vs. (T hi - T ho ) at (m h =0.0497) E.1 (kg/sec) and (m c =0.0414) (kg/sec) Table (E.2) ( ∆ T lm ) vs. (T hi - T ho ) at (m h =0.0579) (kg/sec) and (m c =0.0414) (kg/sec) Table (E.3) ( ∆ T lm ) vs. (T hi - T ho ) at (m h =0.0662) (kg/sec) and (m c =0.0414) (kg/sec) Table (E.4) ( ∆ T lm ) vs. (T hi - T ho ) at (m h =0.0745) (kg/sec) and (m c =0.0414) (kg/sec) Table (E.5) ( ∆ T lm ) vs. (T hi - T ho ) at (m h =0.0828) (kg/sec) and (m c =0.0414) (kg/sec) Table (E.6) ( ∆ T lm ) vs. (T hi - T ho ) at (m h =0.091) (kg/sec) and (m c =0.0414) (kg/sec) Table (E.7) ( ∆ T lm ) vs. (T hi - T ho ) at (m h =0.0993) (kg/sec) and (m c =0.0414) (kg/sec) Table (E.8) ( ∆ T lm ) vs. (T hi - T ho ) at (m h =0.1076) (kg/sec) and (m c =0.0414) (kg/sec) Table (E.9) ( ∆ T lm ) vs. (T hi - T ho ) at (m h =0.1159) (kg/sec) and (m c =0.0414) (kg/sec) Table (E.10) The values of overall heat transfer coefficient (U) as a function of hot water flow rate (m h ) Table (E.11) System parameters for different step change X E.1 E.2 E.2 E.3 E.3 E.4 E.4 E.5 E.5 E.6 List of Figures Figure Page Fig.(1.1) Gasketed plate-and-frame heat exchanger 3 Fig. (1.2) Flow of fluids through a PHE 4 Fig. (1.3) Schematic diagram of PHE 4 Fig. (3.1) Photographic picture of the experimental rig 23 Fig. (3.2) Schematic diagram of the experimental rig 24 Fig. (3.3) Schematic diagram of the cooling tower 27 Fig. (3.4) Schematic diagram of Signal conditioning card (T 1 =T ci , T 2 =T co , T 3 =T hi , T 4 =T ho ) 30 Fig. (3.5) Photographic picture of the interface unit (A- DAQ board, B- Signal conditioning card , C- 30 Relay , D- Power supply) Fig. (3.6) MATLAB simulink used to operate the PHE system (T 1 =T ci , T 2 =T co , T 3 =T hi , T 4 =T ho ) Fig. (4.1) Arrangement of cold and hot streams for PHE (as lumped system) 33 36 Fig. (4.2) (a) Process, (b) Feedback control loop 42 Fig. (4.3) Fuzzy logic control system 48 Fig. (4.4) Biological neuron 52 Fig. (4.5) Basic model of neuron 54 Fig. (4.6) Error back propagation in MLP 58 Fig. (4.7) Neural network training with error backpropagation training algorithm 60 Fig. (4.8) General structure of neural network 62 Fig. (4.9) The block diagram of NARMA-L2 62 Fig. (4.10) The complete controller system with neural network controller NARMA-L2 X 63 Figure Fig. (4.11) NARMA-L2 controller simulink block Fig. (5.1) The relation between overall heat transfer coefficient (U) and hot water flow rate (m h ) Page 63 65 Fig. (5.2) Comparison between experimental and theoretical (T co ) response for +ve different step 67 changes in (m h ) Fig. (5.3.a) The outlet cold water temperature distributions for a counter flow of each plate on PHE Fig. (5.3.b) The outlet hot water temperature distributions for a counter flow of each plate on PHE Fig. (5.4) The final outlet cold water temperature for each plate vs. number of plates in PHE Fig. (5.5) Bode diagram of the PHE Fig. (5.6) Transient response of the PHE with PI controller mode (unit step change) Fig. (5.7) Transient response of the PHE with PID controller mode (unit step change) Fig. (5.8) The comparison between the transient response for PI and PID controllers (unit step change) Fig. (5.9) Simulation model of PHE with classical fuzzy logic controller Fig. (5.10) The comparison between the transient response for PID and classical fuzzy logic controllers Fig. (5.11) Plant identification window Fig. (5.12) Simulation model of PHE with ANN NARMA-L2 controller 68 69 69 72 72 73 73 75 77 79 80 Fig. (5.13) Training of ANN NARMA-L2 controller 80 Fig. (5.14) Testing of ANN NARMA-L2 controller 81 Figure Fig. (5.15) The performance of ANN NARMA-L2 control Fig. (5.16) Transient response of the PHE with ANN NARMA-L2 controller Fig. (5.17) Simulation model of PHE with PID fuzzy controller Fig. (5.18) Transient response of the PHE with PID fuzzy controller Fig. (5.19) The comparison among the transient response for PID, ANN and PID fuzzy controllers Fig. (A.1) Calibration curve of the thermocouple Fig. (A.2) Calibration curve of cold water rotameter Fig. (A.3) Calibration curve of hot water rotameter Page 81 82 83 85 86 A.1 A.1 A.1 Fig. (C.1) (a) Temperature curve for Cohen-Coon tuning. (b) Temperature curve approximation with a C.2 first order dead-time system Fig. (C.2) Definition of gain and phase margins C.3 Fig. (E.1) Temperature difference (T hi - T ho ) as a function of ( ∆ T lm ) for (m h =0.0497) (kg/sec) and E.6 (m c =0.0414) (kg/sec) Fig. (E.2) Temperature difference (T hi - T ho ) as a function of ( ∆ T lm ) for (m h =0.0579) (kg/sec) and E.7 (m c =0.0414) (kg/sec) Fig. (E.3) Temperature difference (T hi - T ho ) as a function of ( ∆ T lm ) for (m h =0.0662) (kg/sec) and (m c =0.0414) (kg/sec) E.7 Figure Page Fig. (E.4) Temperature difference (T hi - T ho ) as a function of ( ∆ T lm ) for (m h =0.0745) (kg/sec) and E.7 (m c =0.0414) (kg/sec) Fig. (E.5) Temperature difference (T hi - T ho ) as a function of ( ∆ T lm ) for (m h =0.0828) (kg/sec) and E.8 (m c =0.0414) (kg/sec) Fig. (E.6) Temperature difference (T hi - T ho ) as a function of ( ∆ T lm ) for (m h =0.091) (kg/sec) and E.8 (m c =0.0414) (kg/sec) Fig. (E.7) Temperature difference (T hi - T ho ) as a function of ( ∆ T lm ) for (m h =0.0993) (kg/sec) and E.8 (m c =0.0414) (kg/sec) Fig. (E.8) Temperature difference (T hi - T ho ) as a function of ( ∆ T lm ) for (m h =0.1076) (kg/sec) and E.9 (m c =0.0414) (kg/sec) Fig. (E.9) Temperature difference (T hi - T ho ) as a function of ( ∆ T lm ) for (m h =0.1159) (kg/sec) and (m c =0.0414) (kg/sec) E.9 Chapter One Introduction 1.1 Introduction Heat exchangers (HEs) are devices that are used to transfer thermal energy between two fluid streams at different temperatures without mixing the two streams. They are one of the most important and frequently used processes in engineering, and one of the thermal components that present nonlinear behavior mainly due to complicated hydrodynamics and temperature dependence of fluid properties. The heat exchange mechanism depends on many other variables such as the heat transfer area, temperature difference, flow rates of the fluids, flow pattern, etc [1-3]. P P HEs are key devices used in a wide variety of thermal applications in the chemical process industries, including petroleum refining and petrochemical processing; in the food industry, for example, for 0T 0T pasteurization of milk and canning of processed foods; in the generation of 0T 0T steam for production of power and electricity; nuclear reaction systems; aircraft and space vehicles; and in the field of cryogenics for the low0T 0T temperature separation of gases. HEs are the workhorses of the entire field of heating, ventilating, air-conditioning, and refrigeration [4, 5]. 0T 0T 0T 0T P P There are several different types of HEs including shell-and-tube, double pipe, plate type and spiral tube. This study is concerned with plate heat exchanger (PHE), which is one of the most common type in practice [6]. P P The first patent for a PHE was granted, in 1878, to Albretch Dracke, a German inventor [7], but the first commercially successful plate-and-frame P P heat exchanger in the world was introduced in 1923 by Dr. Richard Seligman, the founder of the Aluminum Plant and Vessel Company Ltd., commonly known today as APV [8]. Around 1930, the company Alfa Laval, P P Sweden, launched an analogous commercial PHE [7]. P P Chapter One Introduction An PHE is a unit which transfers heat continuously from one media to another media without adding energy to the process [8] P P and the PHE is widely recognized today as the most economical and efficient type of HE on the market [9]. P P On the basis of their specific structure and how the plates are attached together, several types of PHEs are available, the most common type is gasketed PHE. The PHEs consist of a pack of gaskets and corrugated metal plates pressed together with a frame [10, 11] P . A gasket that seals around the plate P prevents fluid mixing. It can also be used to create PHE flow configurations such as series, parallel, and multi-pass arrangements by closing and opening ports at the four plate corners [10]. P P The number of plates, their perforation, the type and position of the gaskets and the location of the inlet and outlet connections at the covers characterize the PHE configuration [11]. P P In the 1930’s PHEs were introduced to meet the hygienic demands of the dairy industry. Today the PHE is universally used in many fields; heating and ventilating, dairy, food processing, pharmaceuticals and fine chemicals, petroleum and chemical industries, power generation, offshore oil and gas production, onboard ships, pulp and paper production, etc [7, 12]. P P Nowadays they are finding increasing usage over wide variety of applications because of the advantages such as flexibility, higher heat transfer, ease of maintenance, compactness, lower rates of fouling, less effect of flow induced vibration and better controllability [13], Plates can be P P easily added or removed depending on the desired application and the equipment is relatively low weight [14]. P P PHEs have been successfully used since the 1930s for single-phase heat transfer from liquid-to-liquid in chemical and food processing industries [15]. P P Chapter One Introduction A typical gasketed PHE is the plate-and-frame heat exchanger. The PHE consists of a pack of corrugated metal plates pressed together into a frame shown in Fig. (1.1). The gaskets between the plates form a series of thin channels where the hot and cold fluids flow and exchange heat through the metal plates. The flow distribution inside the plate pack is defined by the design of the gaskets, the opened and closed ports of the plates and the location of the feed connections at the covers [16, 17]. Appropriate design and P P gasketing permit a stack of plates to be held together by compression bolts joining the end plates. Gaskets prevent leakage to the outside and allow the inter-plate channels to be sealed and to direct the fluids into alternate channels, ensuring the two media never mix. Fig. (1.1) Gasketed plate-and-frame heat exchanger [18]. P P The basic operation of a PHE is similar to any other heat exchanger, in which heat is transferred between two fluid streams through a separating wall. Here, in this case, the separating wall is a plate which is used for heat transfer and to prevent mixing of the streams. As it can be seen from Fig. (1.2) and Fig. (1.3) the hot and cold fluid streams flow into alternate Chapter One Introduction channels between the corrugated plates, entering and leaving via ports at the corner of the plates. Thus, heat transfer takes place from the warm fluid through the separating plate to the colder fluid in a pure counter-current flow arrangement [6]. P P Fig. (1.2) Flow of fluids through a PHE [14]. P P Fig. (1.3) Schematic diagram of PHE [12]. P P 1.2 Control of Heat Exchangers In recent years the performance requirements for process plants have become increasingly difficult to satisfy. Stronger competition, tougher Chapter One Introduction environmental and safety regulations, and rapidly changing economic conditions have been key factors in tightening product quality specifications. A further complication is that modern plants have become more difficult to operate because of the trend toward complex and highly integrated processes. For such plants, it is difficult to prevent disturbances from propagating from one unit to other interconnected units. In view of the increased emphasis placed on safe, efficient plant operation, it is only natural that the subject of process control has become increasingly important in recent years P [19] . Without computer - based P process control systems it would be impossible to operate modern plants safely and profitably while satisfying product quality and environmental requirements. Thus, it is important for chemical engineers to have an understanding of both the theory and practice of process control. The two main subjects related are process dynamics and control. The term process dynamics refers to unsteady-state (or transient) process behavior. Transient operation occurs during important situations such as start-ups and shutdowns, unusual process disturbances, and planned transitions from one product grade to another. The primary objective of process control is to maintain a process at the desired operating conditions, safely and efficiently, while satisfying environmental and product quality requirements. The subject of process control is concerned with how to achieve these goals. Two general approaches to control system design [19]: P P Traditional Approach. The control strategy and control system hardware are selected based on knowledge of the process, experience, and insight. After the control system is installed in the plant, the controller settings are adjusted. This activity is referred to as controller tuning. Chapter One Introduction Model-Based Approach. A dynamic model of the process is first developed that can be helpful in at least three ways: i. It can be used as the basis for model-based controller design methods. ii. The dynamic model can be incorporated directly in the control law. iii. The model can be used in a computer simulation to evaluate alternative control strategies and to determine preliminary values of the controller settings. Several specialized strategies that provide enhanced process control beyond what can be obtained with conventional PID controllers. As processing plants become more and more complex in order to increase efficiency or reduce costs, there are incentives for using such enhancements, which also fall under the general classification of advanced control [19]. P P There are many different control strategies that have been used such as conventional feedback control, cascade control, adaptive control, fuzzy logic (FL) control and artificial neural network (ANN) control [20-24]. P P 1.3 Aim of the Work This work is concerned with dynamic behavior of a PHE and the process control implemented using different control strategies through the following steps: 1. Determining a correlation for the overall heat transfer coefficient of the PHE by finding the effect of the hot water flow rate (m h ) on the R R overall heat transfer coefficient (U) obtained from experimental work. 2. Carrying out the experimental dynamic behavior by measuring the response of the outlet cold water temperature (T co ) under different R R Chapter One Introduction step changes in hot water flow rate (m h ) is compared with the R R simulation results with MATLAB to implement the mathematical model. 3. Justifying the design fitness of the PHE by determining the temperature of the inlet and outlet of each plate using a matrix solution method. 4. Selecting the best control parameters by carrying a tuning procedure using two performance criteria; the integral of the square error (ISE) and integral of the time-weighted absolute error (ITAE). 5. Applying different control strategies such as conventional feedback control, fuzzy logic control and artificial neural network control as well as a comparison among them. Chapter Two Literature Survey 2.1 Introduction Heat exchangers are equipments which transfer the heat from a fluid to another for thermal processes in which two fluids have different temperatures. Plate type heat exchanger (PHE) is the most efficient HE [25]. P P PHEs are important components of process and power industry today. Initially, use of the PHEs was limited to hygienic industries such as food processing, pharmaceuticals and dairy industries primarily due to their ease of clearing. HEs are the subject of many dynamic and control studies. Although considerable effort has been devoted for describing the dynamic behavior of HE, little similar work has been done for PHE. This chapter reviews the literature and studies that deal with dynamics, different control strategies (conventional feedback, fuzzy logic and artificial neural network). 2.2 Dynamic Modeling of Heat Exchanger The objective of the dynamic analysis of the process is to observe how conditions (variables) change with time. The first step in the analysis of a dynamic system is to derive its model. Dynamic process models can be used for simulation studies to get information about the process behavior; the models can also be used for control or optimization studies. Process knowledge may be available as physical relationships or in the form of process data. Dynamic analysis of HEs provides information about transient responses subjected to various disturbances [13]. P P Modeling is the procedure to formulate the dynamic effects of the system that will be considered into mathematical equations. The dynamic behavior can be characterized by the dynamic responses of the system by Chapter Two Literature Survey manipulated inputs and disturbances, taking into account the initial conditions of the system [26]. P P Mathematical models are widely used to design, analyze and control industrial processes. Steady state models are very useful, but for the investigation of start-up and control strategies, the dynamic models are needed. Experimental measurements can be made only of inlet and outlet global temperature of PHE, therefore the temperature profile along a PHE is hardly ever known [27]. P Alwan [28] P P studied the dynamic of PHE using step change technique P applied to cold water flow rate and other variables were maintained almost constant. Recorded outlet cold water temperature analyzed by process reaction curve which shows that the system can be represented as first order with negligible time delay. Time constant was measured for various flow rates and it was concluded that the time constant is inversely proportional to the flow rate. Baker [29] P P studied the dynamic characteristics of a PHE by introducing a sinusoidal disturbance in flow of hot stream through frequency generator, while inlet temperature of cold, hot streams and flow rate of cold stream were maintained almost constant. A theoretical model was proposed based on lumped parameter system. Basic equations were obtained using energy balance and the analysis represented the system as first order whereas considering convective mode of heat transfer resulted in first order lead and second order overdamped lag system for lumped parameter model expressed as: G( ) = S T ( ) = H (τ s +1 ) m ( ) τ s + 2ψ τ s + 1 a CO S 2 h S P 2 p Where: G (s) : transfer function. R R T co(s) : outlet cold water temperature. R R ……….. (2.1) Chapter Two Literature Survey m h(s) : hot water flow rate. R R H τa R τp R : constant. R R Ψ : lead time constant. : lag time constant. : damping coefficient. Various disturbances were introduced to suggest a model for comparison with experimental results. They showed that the transfer function from the final results of experimental and theoretical investigations referred to PHE was of second order (overdamped) system lag and first order system lead with dead time. [30] Khan, et al. P P carried out theoretical and experimental analyses of the dynamic of a counter current flow PHE. They have shown that the transfer function relating the outlet temperature of the cold stream (T co ) and R R the mass flow rate of the hot stream (m h ) were best represented by an R R overdamped second order lag coupled with a first order lead with dead time, namely: G (S ) = T m CO ( S ) = h(S ) ( +1 )e t s τ s + 2ψ τ s + 1 Hτas 2 − ……….. (2.2) d 2 P p They found that the transfer function of theoretical model based on lumped parameter system between (T co ) and (m h ) is a second order R R R R (overdamped) lag combined with a first order lead. Al-Zobai P [31] P conducted a simulation and experimental investigation to study the dynamic of PHE. He used step change to predict the system transient response and he found that the system can be represented by first order lag with dead time. He concluded that the theoretical model based on lumped parameter system is second order (overdamped) system lag with first order system lead. Scariot, et al. P [27] P studied the dynamic behavior of the temperature response curves of the product along the PHE and found the best control parameters by mean of fitting models. The response curves were obtained Chapter Two Literature Survey after step disturbance in the product flow rate, different models were evaluated to identify the dynamic behavior of the control parameters along the PHE. The results were obtained using a simulation codified in MATLAB 6.1 software. The results showed a clear non-linear behavior of the response curves along the PHE. Dwivedi and Das [13] P investigated the transient performance of the P U-type PHE subjected to step flow disturbances. Experiments were executed for various possibilities of step flow transients. They showed that the step change was achieved by changing the hot or the cold flow rate made a difference in response in the transient regime. They suggested the scope of the control system required to regulate the outlet temperatures of the PHEs subjected to dynamic state. Results also indicated the allowable time duration required for the control system to bring back a PHE to a steady state. Thirumarimurugan and Kannadasan P [12] P studied the performance of PHE with different systems. The experimental studies involved the determination of the outlet temperature of both cold and hot fluid for various flow rates for parallel and counter current flow patterns. The waterwater system and other systems were used to determine the performance of type HE. They carried out the comparison between parallel flow and counter current flow HEs. Kapustenko, et al. [32] developed the simplified models for modeling P P of PHE behavior and they improved the temperature control quality of the regulator based on butterfly valve. 2.3 Control of Heat Exchangers The control of HE is complex due to its nonlinear dynamics [33, 34] P and complexity caused by many phenomena such as leakage, friction, P Chapter Two Literature Survey temperature dependent flow properties, contact resistance, unknown fluid properties, etc. 2.3.1 Conventional PI and PID Control The conventional proportional plus integral control (PI) is probably the most commonly used technique [35] P . The PI controller has received a P great deal of attention in the process control areas. It is used as a feed back controller which drives the plant to be controlled with a weighted sum of the error and the integral of that value [36]. P P PID control is one of the earlier control strategies P [37] P and it's the most popular controller used in process control systems due to its remarkable effectiveness and simplicity of implementation. The technique is sufficient for the control of most industrial processes and widely used [38]. P P It needs very little knowledge about the process [39]. P P Many of the studies reported in literature on HE using conventional PI control, can be found in the literature [40-43]. P PID control of HE have P been studied by several researchers [34, 41, 44-55]. P P Alwan [28] studied the conventional controllers of PHE. He applied a P P feedback control loop to the system and concluded that steady state offset of the controlled variable tend to be smaller as the magnitude of flow disturbance gets smaller for setting of proportional action. He applied integral controller to eliminate the offset and getting stable operation. Diaz, et al. P [56] P designed the PI and PID controllers of the HE to observe the performance of the controller and to control the temperature of air passing over it. Also the ANN controller was used for comparison. They found that the PID controller showed better performance than PI and less than neural network control in certain cases. They showed that PI and PID controllers were significantly more oscillated and not able to bring the Chapter Two Literature Survey system to a steady state, but keep the outlet air temperature by adjusting the air speed. Al-Zobai [31] P P conducted a simulation and experimental investigation to study control of PHE. Temperature control had been investigated using PI and PID controller implemented using analog and direct digital control systems. He found the conventional strategy would not be the efficient one. [57] Berto and JR. P P implemented and studied the efficiency of the strategies PID feedback in controlling the pasteurization and the cooling temperatures in a PHE. The controller was to keep those temperatures within the range of ±0.5 oC after disturbances in the product inlet P P temperature occurred. 2.3.2 Computational Intelligence Techniques Traditional control methods have poor performances when applied to industrial processes whose models are strongly non-linear and multivariable -based. Better results can be obtained by applying modern techniques [58]. control P P The computational intelligence (CI) techniques, such as FL and ANNs, have been successfully applied in many scientific researches and engineering practices [59]. P P 2.3.2.1 Fuzzy Logic Control Fuzzy logic can be easily applied to most of applications in industry P [60] . Their great advantage is the possibility to introduce the P knowledge of human experts about proper and correct control of a plant in the controller [61]. P P FL control provides a formal method of translating subjective and imprecise human knowledge into control strategies, thus facilitating better Chapter Two Literature Survey system performance through the exploitation and application of that knowledge [62]. P P Many researches about HEs by means of FL were reported in references [43, 54, 63-67]. P P Skrjanc and Matko [68] P P evaluated the proposed fuzzy predictive control on HE plant, which exhibits a strong nonlinear behavior. It has been shown that in the case of nonlinear processes, the approach using fuzzy predictive control gave very promising results. The main advantage in comparison to the other modern techniques was simplicity together with excellent performance. Al-Zobai [31] P P investigated temperature control using fuzzy control. He obtained a good agreement by experimental responses when using fuzzy control. The fuzzy control algorithm was implemented as a set of rules expressed by conditional statements. He made comparisons among different strategies and the results showed that the fuzzy control can be preferable for control purposes. Chen, et al. [69] designed an exclusive fuzzy control subsystem. They P P studied the fuzzy control means for the supply air temperature of the HE, and accomplished the fuzzy control performance test. According to the their experimental results, the fuzzy control subsystem could not only reduce the testing time for thermodynamic performances of finned-tube heat exchanger; but also actualized easily the stable control of the supply air temperature of the HE. [25] Mastacan, et al. P P used soft computing techniques to control the water temperature of the Alfa Laval type PHE. FL control was implemented and the good performance of the fuzzy control proves that this can be an alternative to the classic control. Maidi, et al. P [42] P explained the proposed design procedure of an optimal PID linear fuzzy controller in general and applied to design a linear Chapter Two Literature Survey (PI-FL) controller that allowed the control of the temperature distribution of the shell and tube heat exchanger. The performance of the fuzzy control system was evaluated by simulation and compared to the conventional PI controller designed optimization. Habbi, et al. [70] first developed a non linear dynamic fuzzy model for P P the HE using a set of input-output observations. The structure in the collected data was determined by means of fuzzy clustering in the inputoutput product space. They designed an efficient fuzzy model-based leak detection algorithm for a pilot heat exchanger. They had proven to be efficient in detecting leaks of different magnitudes in the water circulation pipe. 2.3.2.2 Artificial Neural Network Control ANNs were developed a few decades ago and now widely used in various application areas such as pattern recognition, system identification, and dynamic control. ANN offers a new way to simulate nonlinear, or uncertain, or unknown complex system without requiring any explicit knowledge about input / output relationship [59]. P P Use of artificial intelligence is increasing day by day because of it's adaptability to changes, and ruggedness in control [71]. P P A large number of papers dealing with the artificial neural network for the HE, can be found in the literature [33, 59, 72-81]. P Diaz P [44] P P applied ANNs to the simulation of the steady and dynamic behaviors of HEs, as well as to the control of fluid temperatures. The experiments were carried out in a HE test facility. The ANN predications were obtained using information about the flow rates and inlet temperatures of both fluids in the HE. Numerical tests showed the feasibility of the method and experimental comparison with standard control techniques such as PID proved the ANN to be more accurate. Chapter Two Literature Survey Diaz, et al. [56] extended the ANN technique to control the outlet air P P temperature in HE by changing the air speed. They showed that the present technique performed better than conventional PI and PID control in certain cases when the results were compared with those of standard PI and PID controller. ANN controller was less oscillatory behavior, which allowed the system to reach steady state operating conditions in regions where the PI and PID controllers are not able to perform as well. Diaz, et al. P [2] P investigated the use of adaptive ANNs to control the exit air temperature of HE. They showed that ANNs were a powerful technique to control nonlinear systems. They can be trained to give small errors in prediction and a stable closed-loop feedback control operation. The neuro controller was able to control the experimental facility and adapt to its new conditions for disturbances in the air and water flow rates. It was also able to learn and control the plant behavior for a change in the set point of the temperature. They suggested that ANNs were useful for the control of thermal systems that may change over time. Kharaajoo and Araabi [82] designed a NN based predictive controller P P to govern the dynamics of a heat exchanger pilot plant. HE was a highly non linear process. Advantages of NNs for the process modeling were studied and a NN based predictor was designed, trained and tested as a part of the predictive controller. The dynamics of the plant was identified using a multi layer perceptron (MLP) neural network. The predictive control strategy based on the NN model of the plant was applied then to achieve set point tracking of the output temperature of the plant. Obtained results demonstrated the effectiveness and superiority of the proposed approach. Varshney and Panigrahi P [46] P implemented the NN based control in a LABVIEW platform and compared with the PID control. They investigated experimentally the control of HE in a closed flow air circuit. The temperature inside the test section of the test facility was maintained at a Chapter Two Literature Survey set value by variation of air flow rate over the heat exchanger tube surface and the water flow inside the heat exchanger tubes. They showed that the NN based control had higher speed of response and the steady state error for NN control had a smaller average value than that of the PID control and the control action based on the NN technique less oscillation in comparison to that of the PID based control. Hu, et al. [4] developed the two ANN-based models of a HE. One of P P them was used to predict the steady-state performance of a HE that can be used in practical situations. The other one was used to predict its dynamic performance that can be used in air-conditioning control. They listed experiences in using ANNs, especially those with back-propagation (BP) structures. Also, the weights and biases of our trained-up ANN models. That results showed that NN models were good alternatives to models based on first principles and an actual ANN-based intelligent control will be made possible. Farahani, et al. [83] P dealt with identification and control of a highly P nonlinear real world application and demonstrated the performance and applicability of the proposed methods for an industrial HE. The main difficulties for identification and control of that plant arise from the strongly nonlinear center. ANN based predictive controller using multi layer perceptron (MLP) was designed to govern the dynamics of a HE pilot plant. Using the neuro predictive controller, the outlet liquid temperature of the plant tracked the desired set points by applying the liquid flow rate as a control signal. Biyanto, et al. P [84] P proposed the NN model with nonlinear auto regressive with exogenous input (NARX) structure type multi layer perceptron and developed to describe the complex behavior of a HE in crude preheat train (CPT) . They observed that the developed model had a good predictive capability. The root mean square error (RMSE) between Chapter Two Literature Survey the actual and predicted outlet temperature were found to be less than 0.3 o P C. A model with good predictive capabilities can be used as a tool to P assess the effect of changes in the operating conditions and feed stocks on the performance of the HEs. Selbas, et al. [85] P P applied NN to predict heat transfer rate for PHE. The back-propagation algorithm was used to train and test the network. They were obtained limited experimental data. They showed that the predicted results by NN approach are close to experimental data. NN approach was suitable and simple tool for use in the estimation of heat transfer rates under different operating conditions. The procedure proposed could help the manufacturer and engineers to model HE in engineering applications. Thirumarimurugan and Kannadasan P [12] P used experimental data to develop NNs using general regression neural network (GRNN) model. Those networks were tested with a set of testing data and then the simulated results were compared with actual results of the testing data .They showed that the predicted results are close to experimental data by ANN approach. Vasickaninova, et al. P [34] P studied possibility to use a NN predictive control (NNPC) strategy for control of a HE. The control objective was to keep the output temperature of the heated stream at a desired value and minimize the energy consumption. The NNPC of the HE was compared with classical PID control by simulations experiments. They demonstrated from comparison of the simulation results obtained using NNPC and those obtained by classical PID control the effectiveness and superiority of the NNPC because of smaller consumption of heating medium. Table (2.1) represents the articles of controlling the HE in general and PHE in particular. Chapter Two Literature Survey Table (2.1) represents the articles of controlling the HE. Researchers Year Subject Alwan 1982 Baker 1983 Dynamic behavior and model in PHE Khan 1988 Dynamic behavior and model in PHE Diaz 2000 Skrjanc and Matko 2000 Fuzzy control Diaz and et al. 2001 Adaptive ANN control Diaz and et al. 2001 Dynamic behavior and conventional control in PHE Conventional (PID) control and ANN control Conventional (PI, PID) control and ANN control Dynamic behavior, model, Al-Zobai 2004 conventional (PI, PID) control and fuzzy control in PHE Berto and JR. 2004 PID feedback control Kharaajoo and Araabi 2004 ANN control Hu and et al. 2005 ANN model Scariot and et al. 2005 Dynamic behavior in PHE Varshney and Panigrahi 2005 Chen and et al. 2006 Fuzzy control Farahani and et al. 2006 ANN control Biyanto and et al. 2007 ANN model Dwivedi and Das 2007 Dynamic behavior in PHE Mastacan and et al. 2007 Fuzzy control in PHE Maidi and et al. 2008 Conventional (PID) control and ANN control Conventional (PI) control and fuzzy control Chapter Two Literature Survey Researchers Year Subject Habbi and et al. 2009 Fuzzy model Kapustenko and et al. 2009 Model in PHE Selbas and et al. 2009 ANN model 2009 Steady state and ANN model in PHE Thirumarimurugan and Kannadasan Vasickaninova and et al. 2010 Conventional (PID) control and ANN control Chapter Three Experimental Work 3.1 Introduction This chapter explains and views in details the experimental part of this work. It includes the description of experimental rig and the instrumentation (i.e. computer control system and measuring devices) that are used during this research and also the experimental procedure necessary to generate the corresponding data sets. 3.2 Description of the Experimental Rig The photographic picture and the schematic diagram of the experimental rig used in the present work are appearing in Fig. (3.1) and (3.2), respectively. The main items of the rig are discussed in the following sections: 3.2.1 Plate Heat Exchanger The main part of the experimental rig is a PHE. It was manufactured by APV Company Ltd. England type (JHE) serial No. (1062) and the plates are made of stainless steel with gaskets. It contains 24 corrugated stainless steel plates assembled in counter-current configuration, single pass / single pass for both hot and cold streams. The specifications of the PHE are given in table (3.1) and a listing of experimental rig components appears in table (3.2). Chapter Three Experimental Work Table (3.1) Plate heat exchanger specifications [86]. P Plate length (cm) 58 Plate width (cm) 7 Plate thickness (mm) 1 Equivalent diameter 4 of channel (mm) Channel flow area (m2) 1.4*10-4 Plate pitch (mm) 3 Mean flow channel gap (mm) 2 P P P P The PHE is arranged as U-type flow configuration. Due to the availability and high heat capacity of the water, it was employed in the present work. The use of water as a cooler makes the universal cooling media. 3.2.2 Sump Tank for Hot Water The sump tank has a capacity of hot water (0.07) m3. The tank is P P rectangular with the dimensions of (60*45*30) cm. The sump tank outlet is kept as far away as possible from its inlet to avoid short circuits in the flow. Three immersion heaters were fixed in the sump tank, two immersions heaters were used to achieve required hot stream temperature having a power of 1.5 Kw and 3 Kw respectively and (220-240 V, AC, single phase). Thermostat range on either heater was (40-80) oC, all the pipes work P P which carried hot water were insulated with glass wool. Chapter Three Experimental Work Fig. (3.1) Photographic picture of the experimental rig. Chapter Three Experimental Work Chapter Three Experimental Work Table (3.2) Description of the experimental rig. Code PHE P1 Components Description Plate heat APV plate heat exchanger type (JHE) exchanger serial No. (1062). Cold water 1/2 inch size , 0.37 Kw , 0.3 HP , (0.6- pump 2.4) m3/hr , 32.9 m , 220 V , 60 HZ. P P 1/2 inch size , 0.37 Kw , 0.5 HP , 45 P2 Hot water pump L/min , 35 m , (220~240) V , 50/60 HZ. "GEC Elliot" rotameter , process R1 Cold water instruments Ltd , series 2000 , type rotameter (TM-24 FM-S 1804-V) , range (2-20) L/min. "GEC Elliot" rotameter , process R2 Hot water instruments Ltd , type (TM-14 FM-S rotameter 1802-V) , range (0.5-5) L/min , this rotameter not used in work. R3 Hot water "WDLL" rotameter , Instrument rotameter company , range (2-18) L/min. 1/2 inch size , Siebe Environmental Controls , Loves Park , IL 61111 , MF FCV Flow rate control – 22303 Valve actuator , 24 V, 50/60 valve HZ , 1 watt class 2 , OP. AMB 40 to 140 oF , Temp. Ind. And Reg. Equip. P P and Plenum Rated Cable. Rectangular CT Cooling tower tank for cold water (80*40*55) cm, duct (140*35*30) cm and 12 inch fan. Chapter Three Experimental Work Code Components V1,V7 Description Cold water valve Gate valve , 1/2 inch size. V2,V3,V4, Hot water valve V5,V6 Gate valve, 1/2 inch , the hot water valve V2 not used in work (closed). Rectangular tank (60*45*30) cm with HWT Hot water tank three immersion heaters two in 1.5 Kw and one in 3 Kw. 3.2.3 Cooling Tower The cooling tower was supplied by a manufacturer with a capacity of (0.160) m3. It consisted of a tank with dimensions (80*40*55) cm, duct P P with dimensions (140*35*30) cm and a 12 inch fan. Cold water circulated, and the outcoming (cold) stream from the exchanger was cooled in the cooling tower and sent to the PHE. A cooling tower was used in conjunction with the original rig to ensure availability of the process fluid. Hence a closed loop was established leading to a control over the process fluid inlet temperature to desired levels compatible with the specifications of the rig. The temperature drop obtainable from the cooling tower was 3-9 oC. P P The schematic diagram of the cooling tower and the main dimensions are illustrated in the Fig. (3.3). Chapter Three Experimental Work Fig. (3.3) Schematic diagram of the cooling tower. 3.2.4 Temperature Measurement K-type thermocouples were connected at the entrance and exit pipe lines of both cold and hot sides of the exchanger which able to measure the exchanger temperature response every one second, the details of the K-type thermocouple are illustrated in table (3.3). The responses are recorded with the help of the data acquisition system. All the thermocouples were calibrated before being used to measure the temperatures of the experimental. The apparatus consist of a glass beaker for water bath and a calibrated thermometer. All thermocouples to be calibrated were immersed in a constant temperature bath (cooled water). The temperature in the bath was measured Chapter Three Experimental Work using a precision thermometer. After this, thermocouples calibration can be applicable using the digital multimeter type MASTECH (MS8217) for all thermocouples. The calibration curves are shown in Fig. (A.1) in appendix (A).Similar data were obtained for all the other three thermocouples. Table (3.3) Standardized detail of the K-type thermocouple [87]. P Temperature range (°C) P –250 → 1100 40 from Output (µV/°C) 250–1000°C 35@1300°C Cost Low Stability over the temperature range Low specified Cable specification K Nickel–chromium Common name alloy (chromel)/ nickel–aluminium alloy (alumel) The type K thermocouple is commonly called chromel alumel. It is the most commonly used thermocouple and is designed for use in oxidizing atmospheres. Maximum continuous Brief description use is limited to 1100°C although above 800°C oxidation causes drift and decalibration. Note that the type K thermocouple is unstable with hysteresis between 300°C and 600°C which can result in errors of several degrees. Chapter Three Experimental Work 3.2.5 Water Flow rate Measurement Both hot and cold side rotameters were calibrated before being used in this work to measure the water flow rate of the experiment with standard calibration using calibrated cylinder. The calibration curves are illustrated in Fig. (A.2) and (A.3) in appendix (A). 3.3 Description of the Computer Control System The computer control system required a computer and an interface unit. The interface unit consisted of Data Acquisition Board (NI USB6009), signal conditioning card, relay, and power supply. The computer used to control the system was a personal computer (Pentium four, processor 3.42 GHZ). The interface consisted of an electronic circuit which enables the communication between the system under study and the computer to take advantage of the computer software for generating reports, plots, etc. The schematic diagram of signal conditioning card and photographic picture appears in the Fig. (3.4) and (3.5) respectively. Chapter Three Experimental Work 1 14 1 14 13 2 13 2 13 3 12 3 12 3 12 3 12 4 11 4 11 4 11 4 11 10 5 10 5 10 5 9 6 9 6 9 6 7 8 7 8 7 8 7 8 x x x x x x x x 6 1 out1 2 - 3 + 4 Vcc+ 5 + 6 - 7 out4 LM324 5 out2 - AD595 14 2 AD595 1 13 AD595 14 2 AD595 1 10 9 14 13 12 + Vcc+ 11 10 9 out3 8 T1 T1 T2 T2 T3 T3 T4 T4 Fig. (3.4) Schematic diagram of Signal conditioning card (T1 =T ci , R R R R T 2 =T co , T 3 =T hi , T 4 =T ho ). R R R R R R R R R R R R C A B D Fig. (3.5) Photographic picture of the interface unit (A- DAQ board, B- Signal conditioning card , C- Relay , D- Power supply). Chapter Three Experimental Work The interface unit is consisting of many parts listed below: 1- Data Acquisition Board DAQ (NI USB-6009): The NI USB-6009 is a USB based data acquisition (DAQ) and control device with analog input and output and digital input and output. The main features specifications of NI USB-6009 are as follows: • Analog input (AI): 8 inputs with referenced single ended single coupling. Software-configurable voltage ranges: ±20V, ±10V, ±5V, ±4V, ±2.5V, ±2V, ±1.25V, ±1V. Max sampling rate is 48KS/s (48000 samples per second), and 14 bits AD converter. • Analog output (AO): 2 outputs with voltage range of 0-5V (fixed). The output rate is 150 HZ (samples/second), with 12 bits DA converter. • Digital input (DI) and digital output (DO): 12 channels which can be used as either DI or DO (configured individually). These 12 channels are organized in ports, with port 0 having lines 0,.., 7, and port 1 having lines 0,..,3 . Low input is between -0.3V and +0.8V. High input is between 2.0V and +5.8V. Low output is below 0.8V. High output is above 2V. • Counter: 32 bits. Counting on falling edge. • On-board voltage sources (available at individual terminals): 2.5V and 5.0V. • POWER: USB-6009 is powered via the USB cable. • Application software: LABVIEW, C, or Visual Studio .Platforms: Windows, Mac, or Linux. In this work the DAQ board used with 4 channels analog inputs of 10V, sampling rate was 1000 kS/s and 2 bits digital outputs of 5V. 2- Signal Conditioning Card: This card was built to convert the thermocouple signal to equivalent to (10 mV/oC) by using thermocouple amplifier with cold junction P P Chapter Three Experimental Work compensator. Then this signal was sent to a voltage follower amplifier to increase its loading ability. 3- Relay: The control relay was used to isolate control signal from the loading circuit (AC motor). 4- Power Supply: The power transformer was used to reduce the supplied voltage 220 V AC to 28 V AC . This voltage was supplied to a bridge (Full wave rectifier) R R R R to convert it to DC voltage (i.e. 28 V DC ). Then voltage regulators were used R R to get the required voltages +5V, +24V. P P P P The temperature signal was sent as a voltage difference from a thermocouple to the thermocouple amplifier with cold junction compensator. The process contained four thermocouples. These temperature signals were received from the process by four analog channels in the DAQ Board, which actually has eight channels. The DAQ Board sends these signals as a 14-bit digital number to the computer, where the MATLAB package was used to monitor these temperatures. A manual controller via the MATLAB package was used to set different flow rates values, the generated temperatures due to the different flow rates were read to get the steady state and dynamic data. The manual control action sent from the computer to the DAQ Board. It was sent to two control relays via the DAQ board digital output pins. Two bits were used in this control one to increase the flow rate and the other to decrease it. The control relay was used to isolate control signal from the loading circuit (AC motor). MATLAB simulink was used to operate the PHE system about the DAQ board as illustrated in Fig. (3.6). Chapter Three Experimental Work Fig. (3.6) MATLAB simulink used to operate the PHE system (T 1 =T ci , T 2 =T co , T 3 =T hi , T 4 =T ho ). R R R R R R R R R R R R R R R R 3.4 Experimental Procedure The system was calibrated before every experiment and the data collected from the experiments were of two types: • Steady - state data. • Dynamic response data. 3.4.1 Steady - State Data 1. At the beginning of the experiments, valves V1, V3, V4 and V5 Fig. (3.2) were kept fully open, and V6 and V7 were opened partially to drain the excess water to the tank. 2. The water was pumped by means of electrical pumps both for cold and hot water. 3. After 10 minutes from water circulation in pipes, the valve V4 was closed then the hot water flows in by pass line and back to the hot water tank. Chapter Three Experimental Work 4. The control valve was adjusted to give different flow rates, the hot water flow rate was varied from (0.0497 kg/sec - 0.1159 kg/sec) by the controller via MATLAB package which used to set different flow rates values. 5. The cold water flow rate remained fixed at a constant value of (0.0414 kg/sec) by means of V1. 6. Two electrical heaters were switched on and the thermostat set to 50 oC. P P 7. Valve V5 was closed and the valve V4 was opened to allow the hot water passing through the exchanger. 8. The inlet and the outlet temperatures of fluid streams were recorded by the DAQ board via MATLAB package and the readings were taken at intervals of twenty seconds until steady state readings was reached . 9. The same procedure was repeated for different flow rates of hot water. 3.4.2 Dynamic Response Data 1. Steps 1, 2, and 3 in the previous section were repeated. 2. The control valve was adjusted in order to get the desired flow rate of hot water (0.0497 kg/sec) by means of DAC via MATLAB package. 3. Setting the cold water flow rate to remain fixed at a constant value of (0.0414 kg/sec). 4. The two electrical heaters were switched on and set the thermostat to 50 oC then waiting until the desired temperature was reached. P P 5. Valve V5 was closed then valve V4 was opened. 6. Waiting to reach the steady state by noting that the outlet cold water temperature was fixed via MATLAB program. Chapter Three Experimental Work 7. A step change of (20%) was introduced in hot water flow rate after steady state is reached. 8. The outlet cold water temperature was recorded each five seconds until the new steady state was reached noting that the outlet cold water temperature was fixed by DAQ Board via MATLAB package. 9. The same procedure was repeated for different step changes in hot water flow rate (50%, 80%, 100%, 120%, and 135%). Chapter Four Modeling and Theoretical Analysis 4.1 Introduction In this chapter, the mathematical model for the plate heat exchanger is developed and the control strategies are presented. The theoretical models of various process units are derived by the fundamental principles of conservation of mass and/or energy balance. In its most general form, the conservation state that: Input - Output = Accumulation In steady sate condition, the accumulation is equal to zero. For dynamic simulation the accumulation term to the mass and/or energy balance must be added. Basic equations to present PHE are obtained based upon energy balance. In Fig (4.1) it is considered that each plate operates independently and the transfer function of one plate represents the over all transfer function of the whole PHE. This plate can be considered as lumped system if the theoretical analysis depends upon inlet and outlet temperatures, and variation of temperature along the length is neglected [29, 31]. P Thi mh Cold plate P Tco mc Hot plate Hot plate Cold plate Tho mh Tci mc Fig. (4.1) Arrangement of cold and hot streams for PHE (as lumped system). Chapter Four Modeling and Theoretical Analysis 4.2 Model Assumptions In practice, application of this procedure introduces additional complexity into the system equation, so it is sometimes necessary to make simplifying assumptions of the dynamic behavior. To simplify this complex problem, a few assumptions are made in order to set-up the relevant differential equations. The following assumptions are frequently made in the modeling of PHEs: 1. The physical properties of the water are constant over the range of temperatures employed [29-31]. P P 2. The heat losses to the surroundings are negligible and the two end plates of the exchanger serve as adiabatic walls [29-31]. P P 3. The film coefficient for heat transfer is dependent principally upon the fluid velocity and is proportional to an exponential function of the flow rate [29-31]. P P 4. The heat transfer within the water in any channel is by convection only [29-31]. P P 5. The temperature distributions in all channels belonging to the same stream are identical [29-31]. P P 6. The water will split equally between the parallel channels for each stream [29-31]. P P 7. The thermal capacity of the plate wall is negligible compared with the thermal capacity of the water in the plate [29-31]. P P 8. The plate can be considered as lumped system if the theoretical analysis depends upon inlet and outlet temperatures [29, 31]. P P 9. The variation of temperature along the length is neglected [29, 31]. P P These assumptions are incorporated in the development of a lumped parameter model in which the system may be described by unsteady-state energy balances across any specific plate. Chapter Four Modeling and Theoretical Analysis On considering that the overall heat transfer coefficient (U) is a function of the hot stream mass flow rate m h which in turn is a function of R R time. Hence in the latter instance U is also a function of time, i.e. U (t). 4.3 Energy Balance 4.3.1 Energy Balance around Cold Plate The steady state energy balance around cold plate gives: o o o o+ o + o o dT co = T T hi T ho ci T co − = + − mc C p T ci UA mc C p T co M c C p 0 dt 2 2 o ……….. (4.1) o For dynamic studies the flow rate of hot water is chosen as an input variable while the inlet-temperature of cold and hot streams and flow rate of cold water are maintained constants. The overall heat transfer coefficient is a function of thermal resistance offered by hot and cold stream. As the flow rate of cold stream is considered constant, the overall heat transfer expression is given as: U =α m ……….. (4.2) w h Substituting equation (4.2) into equation (4.1) and simplification takes the shape as: o mc C pT ci − mc C pT co + Z o o o o ow m h T hi + Z o ow m h T ho − Z o ow m h T ci − Z o ow m h dT T co = M c C p co = 0 o dt ……….. (4.3) Where: α Α Z = 2 ……….. (4.4) The unsteady state energy balance around the cold plate gives: m C T − m C T ′ + Z m′ T o c p o o ci c p co w o h hi +Z m′ T ′ − Z m′ T − Z m′ T ′ w h ho w o w h ci h co = M c C p dT ′ co dt ……….. (4.5) The non linear terms in equation (4.5) are linearized using Taylor series: Chapter Four m′ T w o h hi Modeling and Theoretical Analysis = mh ow T + T ∗ w m (m′ − m ) o o w−1 hi hi h ……….. (4.6) o h h Similarly: m′ T ′ w m′ T w o h ci w w−1 o ho h h o ow h h ho T + T ∗ w m (m′ − m ) = mh T + T ∗ w m (m′ − m )+ m (T ′ − T ow co m (m′ − m )+ m (T ′ − T + T ho ∗ w o T = mh ow m′ T ′ h = mh ow ho h o o w−1 ci ci h o o w−1 co co h o ho ) ……….. (4.8) o h h ……….. (4.7) h o ow h h co o co ) ……….. (4.9) Substituting equations (4.6)-(4.9) into equation (4.5) and subtracting the steady-state equation (4.3) from equation (4.5) and introducing deviation variables lead to: τ dT dt c co +T = co K m +K T 1 2 h ……….. (4.10) ho Where: T co m o ……….. (4.11) = m′h − mh ……….. (4.12) = T ′ho − T ho ……….. (4.13) o h o T ho τ = c = T ′co − T co MC m C +Z m c ow c K1 = ……….. (4.14) p o p h Z w m (T + T − T − T m C +Z m w−1 o o o o h hi ho ci co o c ) ……….. (4.15) ow p h ow K2 = Zm m C +Z m ……….. (4.16) h o c ow p h Applying the Laplace transformation: T co ( s ) = K K (1 +τ s ) m ( ) + (1 +τ s )T 1 2 h s c ho ( s ) ……….. (4.17) c The same procedure is repeated around the hot plate. The initial conditions used are listed in table (B.1) in Appendix (B). 4.3.2 Energy Balance around Hot Plate The steady state energy balance around hot plate gives: Chapter Four Modeling and Theoretical Analysis o o − mh C p T hi mh C p T ho + UA o o o o o o+ o + T hi T ho − T ci T co = M C dT ho = 0 h p 2 2 dt ……….. (4.18) Substituting equation (4.2) into equation (4.18) and simplification takes the shape as: o mC T h p − mh C p T ho − Z o o hi o ow o h hi m T −Z ow o h ho m T +Z ow o h ci m T +Z ow o h co m T = M h C p dT ho dt =0 ……….. (4.19) The unsteady state energy balance around hot plate gives: m′ C T h o hi p − m′h C p T ′ho − Z m′ T w o h hi −Z m′ T ′ + Z m′ T w ho h w o h ci +Z m′ T ′ w h co = M h C p dT ho dt ……….. (4.20) The non linear terms in equation (4.20) are linearized using the Taylor series: m′ T ′ ( ) = mh T ho + mh T ′ho − T ho + T ho o ho h o o o o (m′ − m ) o h h ……….. (4.21) Substituting equation (4.21) and equations (4.6)-(4.9) into equation (4.20) and subtracting the steady state equation (4.19) from of equation (4.20) and introducing deviation variables lead to: dT τ dt ho h +T ho = K m +K T 3 4 h ……….. (4.22) co Where: τ h = M C m C +Z m h ow h K3 = ……….. (4.23) p o p [C (T p o hi h ) w m (T + T m C +Z m − T ho − Z o w−1 o o h hi ow ho o h p − T ci − T co o o )] h ow K4 = Zm m C +Z m ……….. (4.25) h o h ……….. (4.24) ow p h Applying the Laplace transformation: T ho ( s ) = K K (1 +τ s ) m ( ) + (1 +τ s )T 3 4 h s h co ( s ) ……….. (4.26) h Substituting T ho(s ) in equation (4.17) into equation (4.26) leads to: Chapter Four G( ) = s T m Modeling and Theoretical Analysis co ( s ) = h(s ) H (1 +τ s ) τ s + 2 ψ τ s +1 ……….. (4.27) a 2 2 p p Where: K +K K 3 K =1 − K K 4 K = 5 1 2 6 = H τ a 2 = K K ψ p ……….. (4.30) 6 K K ……….. (4.31) 1 5 (τ τ ) = c h K = ……….. (4.29) 5 1 τ ……….. (4.28) 1 2 ……….. (4.32) 2 6 (τ +τ ) 2 (τ cτ h) c ……….. (4.33) h 1 2 Thus if U is considered to be a function of t then the resulting transfer function G (S) between T R R co ( s ) and m h(s ) represents a second-order lag with time constant τ p and damping coefficient ψ combined with a first R R order lead element having a time constant τ a . R R 4.4 Control Strategies The main aim of this work is to develop a mathematical model to control the PHE. In this section, the application of conventional feedback control to the PHE is described and discussed. The FL control and ANN control are also described and employed to improve the PHE response. In the present work, MATLAB version 7.10 was used as the simulation software. 4.4.1 Conventional Feedback Control Conventional feedback control in general is the achievement and maintenance of a desired condition by using an actual value of this Chapter Four Modeling and Theoretical Analysis condition and comparing it to a reference value (set point) and using difference between these to eliminate any difference between them. The controller receives a continuous signal of the temperature measurement, which is compared with the set point to produce the actuating signal. The controller thus produces an error signal, which can be used to regulate the control valve. However, the characteristics of the signal produced can be varied to a large extent according to the internal settings of the controller. Fig. (4.2) shows the block diagram of feedback control system. d m process T (a) Controller mechanism d comparator Tsp e(t) + (set point variable) - (error) controller c(t) Tm (measured variable) Final control element m(t) Tout process (Controlled variable) T measuring device (b) Fig. (4.2) (a) Process, (b) Feedback control loop [88]. P P There are three basic types of feedback controllers which described briefly as follow: A- Proportional Controller The output of a proportional controller changes only if the error signals changes. Since a load change requires a new control valve position, the controller must end up with a new error signal. This means that a proportional controller usually gives a steady-state error or offset P [89] . A P proportional controller will have the effect of reducing the rise time but never eliminate the steady-state error [90]. P P Chapter Four Modeling and Theoretical Analysis Since this controller uses the value of error to adjust the input to the process, this type of controller can never fully return the output variable to its set point. This is a disadvantage of proportional action [89]. P P The proportional control action may be described mathematically as [88]: P P C (t ) = K E (t ) + C C ……….. (4.34) S Where: C(t) : controller output. K c : proportional gain of the controller. R R C s : initial value of controller. R R E(t) : error (difference between measured signal and set point). The transfer function for the proportional controller has the form: G (s ) = K C ……….. (4.35) C B- Proportional-Integral (PI) Controller Most control loops use PI controllers. The integral action eliminates steady-state error in temperature. The measured variable can be returned to the set point without excessive oscillation. The smaller the integral time τ I , R R the faster the error is reduced. But the system becomes more underdamped as τ I is reduced. If it is made too small, the loop becomes unstable [89]. R R P P Its actual signal is related to the error by the equation [88]: P C (t ) = K E(t ) + K ∫ E(t ) dt + C t τ C ……….. (4.36) C I P S 0 Where: τ I : integral time constant. R R The PI controller transfer function is: 1 G (s ) = K 1 + τ s C C I ……….. (4.37) The response of the PI controller will be slower than the proportional controller. Thus, the response period of the loop under PI control is 50% Chapter Four Modeling and Theoretical Analysis longer than that for a loop under proportional only control. In order to increase the speed of the response it may be necessary to add an additional control mode [20]. P P C- Proportional-Integral-Derivative (PID) Controller PID controllers are widely used in industrial control systems. The primary purpose of a PID is to provide a fast response that is much the same as with proportional only controller but which has no offset. The derivative action adds the additional response speed required to overcome the lag in the response from the integral action [20]. P P The output of this controller is given by [88]: P t C (t ) = K E (t ) + K ∫ E (t ) dt + K τ dE + C dt τ 0 P ……….. (4.38) C C C D S I Where: τ D : derivative time constant. R R The PID transfer function is given by: 1 + G (s ) = K 1 + τ s τ s C C ……….. (4.39) D I 4.4.2 Controller Tuning Performance of feedback controllers depends on the values of their chosen parameters. If these parameters are properly chosen, they offer the highest flexibility to achieving the desired controlled response and stability. The process of choosing these parameters is known as controller tuning. Also controller tuning can be defined as an optimization process that involves a performance criterion to the form of the controller response and to the error between the process variable and the set point [20]. P P The following two methods are preferred here those of Cohen-Coon method also known as process reaction curve derived from open-loop systems P [91] . The other is Ziegler-Nichols method based on frequency P Chapter Four Modeling and Theoretical Analysis response analysis (bode diagram) [92] P P that it derived mainly from closed- loop systems. The two methods are described in appendix (C). The main two methods of the time integral performance criteria used in the proposed work evaluated in terms of: Integrated Square Error (ISE) This error uses the square of the error, thereby penalizing large errors more than small errors. This gives more conservative response (faster return to set point) [20]. P P ∞ ……….. (4.40) ISE = ∫ e dt 2 0 Integrated Time-Weighted Absolute Error (ITAE) This criterion is based on the integral of the absolute value of the error multiplied by time. It results in errors existing over time being penalized even though may be small, which results in a more heavily damped response [20]. P P ∞ ITAE = ∫t ……….. (4.41) e dt 0 If the performance indices increases, control system can perform poorly and even become unstable. So it needs to tune the controller parameters to achieve good control performance with the proper choice of tuning constants [90]. P P 4.4.3 Fuzzy Logic Control 4.4.3.1 Introduction of Fuzzy Logic FL is one of the tools of what is commonly known as computational intelligence (CI) and it's a logical system, which is an extension and generalization of multi valued logic systems [93]. P P FL is much closed in spirit to human thinking and natural language than classical logical systems. Nowadays FL is used in almost all sectors of industry and science [35, 94, 95]. P P Chapter Four Modeling and Theoretical Analysis FL is one of the successful applications of fuzzy set in which the variables are linguistic rather than the numeric variables [96]. P P In FL theory, the range of values for a given input or output space is often called the universe of discourse. For greater flexibility in fuzzy controller implementation, the universes of discourse are “normalized” to a certain interval (e.g., [-1, 1] or [0, 1]) by means of constant scaling factors [93]. P P FL starts with the concept of a fuzzy set. The concept of fuzzy set theory was introduced by Zadeh in 1965. Fuzzy set theory can be considered as a generalization of the classical set theory. In classical set theory an element of the universe either belongs to or does not belong to the set. Thus the degree of association of an element is crisp. In a fuzzy set theory the association of an element can be continuously varying. Mathematically, a fuzzy set is a mapping (known as membership function) from the universe of discourse to the closed interval [0, 1]. The membership function is usually designed by taking into consideration the requirement and constraints of the problem. FL implements human experiences and preferences via membership functions and fuzzy rules. Due to the use of fuzzy variables, the system can be made understandable to a non-expert operator. In this way, FL can be used as a general methodology to incorporate knowledge, heuristics or theory into controllers and decision makers [95, 97]. P P 4.4.3.2 Linguistic Variables The concept of a linguistic variable, a term which is used to describe the inputs and outputs of the FL control, is the foundation of FL control systems. A conventional variable is numerical and precise. It is not capable of supporting the vagueness in fuzzy set theory. By definition, a linguistic variable is made up of words, sentences or artificial language which is less Chapter Four precise than Modeling and Theoretical Analysis numbers. It provides the means of approximate characterisation of complex or ill-defined phenomena. A more common example in fuzzy control would be the linguistic variable ‘ERROR’, which may have linguistic values such as ‘POSITIVE’, ‘ZERO’ and ‘NEGATIVE’. The following conventions are used to define linguistic variables P [98] . If X i is a linguistic variable defined over the universe of P R R discourse U where x ∈ U then LX i k (for k = 1, . . . n) are the linguistic values X i can take. n is the number of linguistic values X i have. R RP P R R R R μ LXi,k (x) is the LX i k membership function for the value x. LX i is the set containing LX i k , where LX i = { LX i 1, LX i 2 … LX i n }. R R R R R RP P R RP P R R R RP P R RP P R RP P In the example above: X1 is ‘ERROR’ n=3 is the number of linguistic values in X 1 LX 1 1 is ‘POSITIVE’ R RP P LX 1 2 R R is ‘ZERO’ RP LX 1 3 RP R P is ‘NEGATIVE’ P And, for x = {–1, 0, 1}: μ LX1,1 (–1) = 0; μ LX1,1 (0) = 0; μ LX1,1 (1) = 1 μ LX1,2 (–1) = 0; μ LX1,2 (0) = 1; μ LX1,2 (1) = 0 μ LX1,3 (–1) = 1; μ LX1,3 (0) = 0; μ LX1,3 (1) = 0 R R R R R R R R R R R R R R R R R R 4.4.3.3 Fuzzy Logic Controller The idea of FL controller was initially introduced by Zadeh (1973) and first applied by Mamdani (1974) in an attempt to control systems that are difficult to model mathematically [93]. P P The Mamdani fuzzy model is the first working model of fuzzy control systems. It constructs a bridge between the operator's knowledge Chapter Four Modeling and Theoretical Analysis and IF-THEN rules by fuzzy logic. The Mamdani fuzzy model provides a basic procedure for fuzzy controller design [99]. P P 4.4.3.3.1 Design of Fuzzy Logic Controller An alternative approach to design a controller as compared to transfer function based controller is to use linguistic control protocol employed by a human operator as used in FL controller. A block diagram of FL control system is shown in Fig. (4.3), which comprises of a fuzzification interface, data base, rule base, inference mechanism and a defuzzification interface [94, 100-103] P . Sometimes both data and rule base can P be called as knowledge base [94, 100]. P P Fig. (4.3) Fuzzy logic control system [104]. P P The fuzzification interface measures the values of input variables, performs a scale mapping that transfers the range of values of input variables, and converts input data into suitable linguistic values [94]. P P This transformation is performed using membership functions. In a FL controller, the number of membership functions and the shapes of these Chapter Four Modeling and Theoretical Analysis are initially determined by the user. The membership functions can take many forms including triangular, gaussian, trapezoidal, etc [105]. P P The data base provides necessary definitions, which are used to define linguistic control rules and fuzzy data manipulation in an FL control [94]. P P The rule base characterizes the control goals and control policy of the domain experts by means of a set of linguistic control rules P [94] . The relationship between input and output variables are P described in a rule base composed of IF-THEN form rules. Usually fuzzy systems are synthesized using two types of rules that differ in the consequent (Then part) proposition form: Mamdani, or standard and Takagi-Sugeno, or functional [54]. P P The main part of the FL controller is the rule base and the inference mechanism. The rule base is normally expressed in a set of fuzzy linguistic rules, with each rule triggered with varying belief for support. The ith linguistic control rule can be expressed as: R i : IF e is A i and de is B i THEN u is C i , R R R R R R R R Where A i and B i (antecedent), C i (consequent) are fuzzy variables R R R R R R characterized by fuzzy membership functions [100]. P P For FL controller the error (e) and the change of error (de) as input linguistic variables are taking and the control action (u) taken as output linguistic variable. For example, the membership function for error, change of error and control action consist of negative (N), zero (Z), positive (P). The set of fuzzy rule for FL control can be written in a table as shown in table (4.1). Chapter Four Modeling and Theoretical Analysis Table (4.1) IF-THEN rule base for fuzzy logic control. de N Z P N P P Z Z P Z N P Z N N e The inference mechanism (also called an “inference engine” or “fuzzy inference” module), which emulates the expert’s decision making in interpreting and applying knowledge about how best to control the plant [104]. P P The decision making logic is the kernel of an FL controller, and has the capability of simulating human decision-making based on fuzzy concepts [94]. P P The defuzzification interface converts the range of values of output variables into corresponding universe of discourse, and yields a nonfuzzy control action from a fuzzy control action [94]. P P The performance of the FL control depends very much on the defuzzification process. This is because the overall performance of the system under control is determined by the controlling signal (the defuzzified output of the FL controller) that the system universe [100]. P P The design of a fuzzy controller remains a difficult task due to the fact that there is insufficient analytical design technique in contrast with the well-developed linear control theories. Although the fuzzy controller has the advantage of being relatively easy to understand, the controller tuning is complex or nontransparent due to many involved design parameters, and in most cases the fuzzy controller design is accomplished by trial and error methods using computer simulations. It is an attempt to undertake the Chapter Four Modeling and Theoretical Analysis development of an approach to the optimal design of linear PID fuzzy controller [42]. P P 4.4.4 Artificial Neural Network Control 4.4.4.1 Introduction of Artificial Neural Network An ANN takes their name from the network of nerve cells in the brain [106] and it's a flexible mathematical structure, having strong similarity P P to the biological brain and therefore a great deal of the terminology is borrowed from neuroscience [107]. P P An ANN is an information-processing paradigm that is inspired by the way biological nervous systems, such as the brain, process information. It is composed of a large number of highly interconnected processing elements (neurons) working in unison to solve specific problems [33, 108, 109] P P and they are powerful tools that can learn to solve problems in a way similar to the human brain. ANNs gather knowledge by detecting the patterns and relationships in data and learn (or: are trained) through experience, not from programming [110] especially when the underlying data P P relationship is unknown [111]. P P 4.4.4.2 Biological Artificial Neural Network An ANN is an information processing system that has been developed as a generalization of the mathematical model of human cognition. A biological neuron has three types of components, namely, the dendrites, soma and axon. Dendrites are bunched into highly complex “dendritic tree”, which has an enormous total surface area. The dendrites receive signals from other neurons. Dentritic trees are connected with the main body of the neuron called the soma. The soma has a pyramidal or cylindrical shape. The soma sums the incoming signals. The cell fires when Chapter Four Modeling and Theoretical Analysis sufficient input is received. The output area of the neuron is a long fibre called the axon. The impulse signal triggered by the cell is transmitted over the axon to other cells. The connecting point between a neuron’s axon and another neuron’s dendrite is called a synapse. The impulse signals are transmitted across a synaptic gap by means of a chemical process. A single neuron may have 1000 to 10000 synapses and may be connected with 1000 neurons. There are 100 billion neurons in the brain and each neuron has 1000 dendrites [112, 113]. Fig. (4.4) shows the biological neuron. P P Fig. (4.4) Biological neuron [112]. P P 4.4.4.3 Mathematical Model of a Neuron A first wave of interest in ANNs (also known as connectionist models or parallel distributed processing) emerged after the introduction of simplified neurons by McCulloch and Pitts (1943) [114]. P P ANNs are inspired by the early models of sensory processing by the brain. An ANN can be created by simulating a network of model neurons in a computer. By applying algorithms that mimic the processes of real neurons, we can make the network ‘learn’ to solve many types of problems [115]. P P Chapter Four Modeling and Theoretical Analysis The artificial neuron imitates the characteristics of the biological neuron. A processing element possesses a local memory and carries out localized information processing operations. The artificial neuron has a set of ‘p’ inputs x i , each representing the output of another neuron (the R R subscript i takes values between 0 and p and indicates the source of the vector input signal). The inputs are collectively referred to as X. Each input is weighted before it reaches the main body of the processing element by the connection strength or the weight factor analogous to the synaptic strength. The amount of information about the input that is required to solve a problem is stored in the form of weights. Each signal is multiplied with an associated weight wk 1 , wk 2 ,…, wk p before it is applied to the R R R R R R summing block. Equation (4.42) and (4.43) show how the sum of the weights is calculated. In addition, the artificial neuron has a bias term, wk 0 , R R a threshold value that has to be reached for the neuron to produce a signal. Activation function is needed to produce the output, y k as shown in R R equation (4.44). The basic model of a neuron is shown in Fig. (4.5). It should be noted that the input to the bias neuron is assumed to be 1.The main task of the activation function is to map the outlying values of the obtained neural input back to a bounded interval such as [0, 1] or [-1, 1]. The transfer function of the basic neuron model is described [112]. P v = x wk + x wk + x wk + x wk + ........ + x wk k o o 1 1 v = wk + ∑ x wk 2 p k o y = F (v k ) k i =0 i i 2 3 3 p p P …….... (4.42) .…..…. (4.43) ………. (4.44) Chapter Four Modeling and Theoretical Analysis Fixed Fixed Input X o = ± 1 X o= ± 1 Input XX o wk = wk o wk o = b(bias) b k (bias) wk o o XX o k wk 1 wk 11 1 Activation Activation Function Function XX 2 wk wk 2 2 2 ∑∑ v vk k Output Output ϕ( ϕ ) ( ) o y y k o k Summing Summing Junction Junction X X Input wk P P wk P P Synaptic Synaptic Weight Weight Input signals signals Fig. (4.5) Basic model of neuron. A common choice for the threshold function is the sigmoid activation function: ϕ (x ) = 1 1 + e cv − ………. (4.45) k Where c is a constant parameter that determines the shape of the sigmoid [116]. P P Sigmoid function is used for the activation function due to some of its advantages [117]: P P 1. Nonlinearity makes the learning powerful. 2. Differential is possible and easy with simple equations. 3. Negative and positive value makes learning fast. 4.4.4.4 Architecture of Artificial Neural Network There are several types of ANNs according to their structure (or Architecture) and learning algorithms. According to their structure, ANNs can be classified as feed forward networks and recurrent networks [110]. P P Chapter Four Modeling and Theoretical Analysis The most common for chemical engineering application is Multi Layer Perceptron (MLP), which is a feed forward neural network. An MLP is a powerful system, often capable of modeling complex, relationships between variables. It allows prediction of an output object for a given input object. It consists of multilayer structure, which a part from input and output layers, has at least one layer of processing units in between them. The layers between the input and output layers are termed "hidden" since they do not converse with the outside world directly. The nodes between the two successive layers are fully connected by means of weights. That is outputs from the input layer are fed to the hidden layer units, which in turn, feed their outputs to the next hidden nodes. The hidden node passes the net activation through a nonlinear transformation of a linear function, such as the logistic sigmoidal to compute their outputs [33]. P P 4.4.4.5 Back Propagation (BP) Algorithm Artificial Neural Network An ANN starts with a set of initial weights and then gradually modifies the weights during the training cycle to settle down to a set of weights capable of realizing the input-output mapping with either no error or a minimum error set by the user [118]. P P For many years, there was no theoretically sound algorithm for training multilayer ANNs. The invention of the BP algorithm has played a large part in the resurgence of interest in ANNs. BP is a systematic method for training multilayer ANNs (Perceptrons). The BP learning algorithm is currently the most popular learning rule for performing supervised learning tasks [119]. P P A learning algorithm of an MLP is called a Multi-Layer Back Propagation (MLBP). Chapter Four Modeling and Theoretical Analysis The main idea of this algorithm is to minimize cost function by steepest descent method to add small changes in the direction of minimization [59]. P P The feed forward back propagation network is a very popular model in neural networks. It does not have feedback connections, but errors propagate backward from the output layer during training. BP is a gradient descent method to minimize the total squared error of the output computed by the NN [120]. P P It has been proven that BP learning with sufficient hidden layers can approximate any nonlinear function to arbitrary accuracy. This makes back propagation learning neural network a good candidate for signal prediction and system modeling [114]. P P The back propagation learning algorithm is performed in the following steps [114,117]: P P 1. Initialize network weight values. 2. Repeat the following steps until some criterion is reached: (for each training pair). 3. Sums weighted input and apply activation function to compute output of hidden layer. h i = f (∑ X W i i ij ) ………. (4.46) 4. Sums weighted output of hidden layer and apply activation function to compute output of output layer. y K = f (∑ h W j j jK ) ………. (4.47) 5. Compute back propagation error. δ K = (d K − y ) f ′ (∑ h W j K j jK ) ………. (4.48) 6. Calculate weight correction term. ∆W (n) = η δ h + α ∆W (n −1) jK K j jK ………. (4.49) 7. Sums delta input for each hidden unit and calculate error term. Chapter Four Modeling and Theoretical Analysis δ = ∑ δ W f ′ (∑ X W ) j K K jK i i ij ………. (4.50) 8. Calculate weight correction term. ∆W (n) = η δ X + α ∆W (n −1) ij j i ………. (4.51) ij 9. Update weights. W (new) = W (old ) + ∆W jK jK W (new) = W (old ) + ∆W ij ij ………. (4.52) jK ………. (4.53) ij 10. End. These steps are illustrated in Fig. (4.6). Where: ΔWij : Amount of Change Added to The Weight Connection W ij . R R y K : Output Signal of an Output Neuron (K) at Time (n). R R d K : Desired (Target) Output Neuron (K) at Time (n). R R η : Learning Rate Coefficient. h j : Output Signal of Hidden Neuron (j) at Time (n). R R δ j : Delta Quantity for Hidden Neuron (j). R R δ K : Delta Quantity for Output Neuron (K). R R α : Momentum Constant. R R Chapter Four Modeling and Theoretical Analysis Inputs X X 1 X 2 X 3 Forward Pass Back Pass X i W (n) = W (n−1) + ∆W (n) W ij i ij h h 1 h 2 ij y 1 ij ) j Y y j i W K j i j ij δ = ∑ δ W f ′ (∑ X W j K K jK i i ij ) W (n) = W (n−1) + ∆W (n) jK jK f (∑ h W = ij ∆W (n) =η δ X + α ∆W (n−1) h = f (∑ X W j ij j jK ) jK jK ∆W (n) =η δ h + α ∆W (n−1) jK K j jK δ = (d − Y ) f ′ (∑ h W K K K K j j jK ) Outputs Fig. (4.6) Error back propagation in MLP. 4.4.4.6 Artificial Neural Network Controller As in real world of control engineering the nonlinearities are an unavoidable problem that necessitates the development of controllers with special capabilities in solving the nonlinearity problems. ANNs have been proved a successful method in identification and control of dynamic systems. Their approximation capabilities of MLP made them a popular choice for modeling nonlinear systems and for implementing general – purpose nonlinear controllers [121]. P P Different control algorithms and architectures are implemented. One of them, among others, for prediction and control is the NARMA–L2 (or feedback linearization) controller. NARMA-L2 control is one of the three popular ANN architectures that have been implemented in the neural network toolbox of MATLAB software. Nonlinear auto regressive-moving average (NARMA-L2) Chapter Four Modeling and Theoretical Analysis controller is simply a rearrangement of the ANN plant model, which is trained offline in batch form, so it is the best technique to be presented in this work [122-124]. This ANN controller is referred as feedback linearization P P when the plant model has a particular form (companion form) and it is referred to as NARMA-L2 control when the plant model can be approximated by the same form. The central idea of this type of control is to transform nonlinear system dynamics into linear dynamics by canceling the nonlinearities P [125] . The drawback of this method is that the plant must P either be in companion form, or be capable of approximation by a companion form model. NARMA-L2 controller is implemented as simulink block, which is contained in the neural network toolbox blockset. 4.4.4.6.1 Identification and Controller Stages of the NARMA-L2 model NARMA-L2 controller, a multilayer neural network has been successfully applied in the identification and control of dynamic systems. System identification and control design are the two steps involved in using NARMA-L2 controller. The identification of the system by this controller can be summarized by the following steps [122, 125, 126]: P P Identify the system to be controlled. A neural network of the plant that needs to be controlled is developed. One standard model that has been used to represent general discrete-time nonlinear systems is the NARMA-L2 model: y (k + d ) = N [ y (k ), y (k − 1),...., y (k − n + 1), u (k ), u (k − 1),...., u (k − n + 1)] ………. (4.54) Where u(k) is the system input, y(k) is the system output and k,d,n are integral number. Fig. (4.7) shows the block diagram representation of the system identification stage, where the NN training with error backpropagation training algorithm. Chapter Four Modeling and Theoretical Analysis Fig. (4.7) Neural network training with error back-propagation training algorithm. Make the output system follows some reference trajectory by developing a nonlinear controller of the form y(k + d ) = y (k + d ) ………. (4.55) r u(k ) = G y(k ), y(k −1),...., y(k −n +1), y (k + d ) ,u(k −1),...,u(k −m+1) r ………. (4.56) To implement the controller model with NARMA-L2, one solution is to use approximate models to represent the system: yˆ (k + d ) = f [ y (k ), y (k − 1),...., y (k − n + 1), u (k − 1),...., u (k − m + 1)] + g [ y (k ), y (k − 1),...., y (k − n + 1), u (k − 1),...., u (k − m + 1)].u (k ) ………. (4.57) Where f(.) and g(.) are approximated using NNs. The advantage of this form is that one can obtain the controlled input that makes the system output follows the reference in equation (4.55). Using this NARMA-L2 model, the resulting controller has the form: Chapter Four u(k ) = Modeling and Theoretical Analysis y (k + d ) − f [y(k ), y(k −1),..., y(k − n +1),u(k −1),....,u(k − n +1)] g[y(k ),..., y(k − n +1),u(k −1),....,u(k − n +1)] r ………. (4.58) Using this equation directly can cause realization problems, because must determine the control input based on the output at the same time , i.e. y (k + d ) = f [ y (k ), y (k − 1),...., y (k − n + 1), u (k ), u (k − 1),...., u (k − n + 1)] + g [ y (k ),..., y (k − n + 1), u (k ),..., u (k − n + 1)]u (k + 1) .………. (4.59) Fig. (4.8) shows the structure of NN representation of equation (4.59). The block diagram of NARMA-L2 controller together with the reference model and the plant is shown in Fig. (4.9) and the controller which make (e c ) [the difference between plant output (y) and the reference R R y(r)] very small by evaluated input plant (u). The controller system that can be implemented with the previously identified NARMA-L2 plant model is shown in Fig. (4.10). The NARMA-L2 controller simulink block which is contained in the neural network toolbox blockset is shown in Fig. (4.11). Chapter Four Modeling and Theoretical Analysis Fig. (4.8) General structure of neural network. Fig. (4.9) The block diagram of NARMA-L2. Chapter Four Modeling and Theoretical Analysis Fig. (4.10) The complete controller system with neural network controller NARMA-L2. Fig. (4.11) NARMA-L2 controller simulink block. Chapter Five Results and Discussion 5.1 Introduction This chapter presents the results obtained from the computer programs using MATLAB program version 7.10 cited in appendix (D) for dynamic model and control. The first part of this chapter shows the results of the open loop experimental and theoretical response for different magnitudes of step change in hot water flow rate. The second part shows the results of the control system using different control strategies. 5.2 Open Loop System 5.2.1 Steady State Results Tables (E.1) to (E.9) in appendix (E) include the data of ( ∆ T lm ) and R R temperature difference (T hi - T ho ) obtained from steady state data to R R R R calculate the overall heat transfer coefficient for the conditions of counter current flow for different vales of hot water flow rate (0.04970.1159)(kg/sec) while the cold water flow rate remained constant at (0.0414)(kg/sec), using equation (F.7) in appendix (F). Figs. (E.1) to (E.9) in appendix (E) show the relations between ( ∆ T lm ) vs. (T hi - T ho ). The results show the linear fitting represented by R R R R R R equation (F.7) of the experimental data which show a good fitting with little deviations according to the heat losses to the surrounding although efforts had been spent to insulate the exchanger system. From these figures the values of the overall heat transfer coefficient were determined and presented in table (E.10) in appendix (E). In Fig. (5.1) the values of overall heat transfer coefficient (U) are plotted against hot water flow rate (m h ), R R from which the following correlation has been obtained using the method explained in appendix (F). U = 11045 m 0.7158 h ……….. (5.1) Chapter Five Results and Discussion Since the change in the physical properties of water is negligibly small for the same flow rate, the above correlation can be applied for both cold and hot streams and can be written as: ……….. (5.2) U = 11045 m 0.7158 The above correlation appears to be in good agreement with those reported by literature [128,129]. P P Previous reports show that the correlation has the form: h∝m ……….. (5.3) w Where w ranges between (0.697-0.864). In order to present a comparison between the present model equation (5.2) and the above model, equation (5.3) may be written as: h =α m ……….. (5.4) w Since: U = f (h) ……….. (5.5) ∴U = α m ……….. (5.6) w Or: ……….. (5.7) U =α m 0.697 − 0.864 Comparing equation (5.7) with equation (5.2), the agreement between the two models is quite apparent. Fig. (5.1) The relation between overall heat transfer coefficient (U) and hot water flow rate (m h ). R R Chapter Five Results and Discussion 5.2.2 Dynamic Behavior The dynamic responses were studied for different step changes in the manipulated variable (m h ) in order to study the effect of each change on R R the controlled variable (T co ). These changes are: R R +Ve (20%, 50%, 80%, 100%, 120% and 135%) step changes in the hot water flow rate (m h ). R R The theoretical results are obtained by using computer simulation programs given in (D.2) in appendix (D). The theoretical results are compared with experimental results for different step changes. Fig. (5.2) represents the comparison between experimental and theoretical outlet cold water temperature response. Table (5.1) illustrates the relative error (Er) and mean square error (MSE) between experimental and theoretical responses. Fig. (5.2) shows the relation between controlled variable (T co ) and R R time. It can be seen that the increase in hot water flow rate (m h ) is directly R R proportional to outlet cold water temperature (T co ) for different steps in the R R hot water flow rate (m h ) for theoretical and experimental work. R R Table (5.1) Shows that the theoretical results are in good agreement with experimental results. The determined values of the steady-state gain (K) and the time constant (τ) experimentally by using process reaction curve are tabulated in table (E.11) in appendix (E). The analysis indicated that the process can be experimentally represented by first order. Chapter Five Results and Discussion 20 50 80 33 Outlet Cold Water Temperature (C) 100 120 32 135 exp 31 30 29 28 27 0 20 40 60 80 100 Time (sec) 120 140 160 180 Fig. (5.2) Comparison between experimental and theoretical (T co ) R response for +ve different step changes in (m h ). R R Table (5.1) The relative error (Er) and mean square error (MSE) between experimental and theoretical (T co ) response. R R mh Er MSE 20% 0.1110 3.3329e-004 50% 5.2986 0.7588 80% 0.9694 0.0254 100% 0.8960 0.0217 120% 5.8441 0.9231 135% 1.8257 0.0901 R R Chapter Five Results and Discussion 5.2.3 Justifying the Design Fitness of the PHE The design fitness of the PHE has been justified by determining the inlet and outlet temperature of each plate using a matrix solution method (Gaussian-Elimination) [130]. P P The value of hot water flow rate (m h ) used in the design is R R (0.0993)(kg/sec) and the value of cold water flow rate (m c ) is R R (0.0414)(kg/sec). Figs. (5.3.a) and (5.3.b) represents the outlet cold and hot water temperature distributions for a counter flow of each plate on PHE respectively and Fig. (5.4) represents the final outlet cold water temperature for each plate vs. number of plates in PHE. Figs. (5.3) and (5.4) show the response of 24 plates which prove that the design is accurate and there is no losses in the energy input. This is clear through the exponential shape of Fig. (5.4) which proves the reasonable number of plates and accurate geometric design of each plate. 34 Outlet Cold Water Temperatures (C) 32 30 28 26 24 22 20 0 Plate 1 Plate 24 10 20 30 40 50 60 Time (sec) 70 80 90 100 Fig. (5.3.a) The outlet cold water temperature distributions for a counter flow of each plate on PHE. Chapter Five Results and Discussion 50 Plate 1 Plate 24 Outlet Hot Water Temperatures (C) 48 46 44 42 40 38 0 10 20 30 40 50 60 Time (sec) 70 80 90 100 Fig. (5.3.b) The outlet hot water temperature distributions for a counter flow of each plate on PHE. Outlet Cold Water Temperature (C) 34 33 32 31 30 29 Plate 0 5 10 15 Number of Plates 20 25 Fig. (5.4) The final outlet cold water temperature for each plate vs. number of plates in PHE. Chapter Five Results and Discussion 5.3 Closed Loop System In this section, different control strategies are used: conventional feedback control, classical FL control, ANN control and PID fuzzy control. The Value of hot water flow rate used in the control system is (0.0993) (kg/sec) and the value of cold water flow rate is (0.0414) (kg/sec). 5.3.1 Conventional Feedback Control Conventional feedback control was applied using PI and PID modes to control the outlet cold water temperature. The tuning of the control parameters (proportional gain (k c ), time integral (τ I ) and time derivative R R R R (τ D )) were applied. R R The optimum values of the controller parameters (k c , τ I , τ D ) were R R R R R R tuned by using computer simulation programs based on minimum integral of square error (ISE) and minimum integral of the time-weighted absolute error (ITAE). These programs are shown in programs (D.3.1) and (D.3.2) in appendix (D). The results of the control tuning parameters are given in tables (5.2) and (5.3). To evaluate the performance of the PI and PID controllers we have considered three parameters of the step response and the parameters (overshoot, settling time and rise time) have been given in the table (5.4). Table (5.2) Control parameters of PI control. Control tuning 0B methods Controller parameters Kc τI τD ITAE ISE 10.6614 0.1587 ــــ 0.2163 0.2517 10.8959 0.1596 ــــ 0.2795 0.2804 R R Ziegler-Nichols tuning Cohen-Coon tuning R Chapter Five Results and Discussion Table (5.3) Control parameters of PID control. Control tuning 1B methods Controller parameters Kc τI τD ITAE ISE 13.7972 0.0952 0.0238 0.0298 0.1236 16.1441 0.1181 0.0175 0.1871 0.2705 R R R Ziegler-Nichols tuning Cohen-Coon tuning Table (5.4) Comparison of different parameters of PI and PID controllers. Parameters PI controller PID controller Overshoot 2.177 2.12 Settling time 1.61 0.73 Rise time 0.086 0.077 As shown in the tables (5.2) and (5.3) the control tuning was found in two different methods therefore; it can be seen that the tuning by using the Ziegler-Nichols method is better than that of Cohen-Coon method because Ziegler-Nichols method depends on closed loop system while Cohen-Coon method depends on open loop system, also the ISE values and ITAE values of first method are less than that of the second method. In this work the ITAE is implemented because it uses the time to determine its value which states the faster criteria to reach the new steadystate value. As shown in table (5.4) it is clear that PID controller is better than PI controller because it gives smaller overshoot, settling time, and rise time values than that of PI controller. Chapter Five Results and Discussion 5.3.1.1 Control Behavior Bode Diagram 40 Magnitude (dB) 20 0 -20 -40 -60 360 Phase (deg) 270 180 90 0 -90 -3 10 -1 -2 10 10 0 10 2 1 10 10 3 10 Frequency (rad/sec) Fig. (5.5) Bode diagram of the PHE. Ziegler-Nichols Method Cohen-Coon Method Outlet Cold Water Temperature (C) 2 1.5 1 0.5 0 0 0.5 1 1.5 2 Time (sec) 2.5 3 3.5 Fig. (5.6) Transient response of the PHE with PI controller mode (unit step change). Chapter Five Results and Discussion Ziegler-Nichols Method Cohen-Coon Method Outlet Cold Water Temperature (C) 2 1.5 1 0.5 0 0 0.5 1 2 1.5 Time (sec) 2.5 3 3.5 Fig. (5.7) Transient response of the PHE with PID controller mode (unit step change). PID Controller PI Controller Outlet Cold Water Temperature (C) 2 1.5 1 0.5 0 0 0.5 1 1.5 2 Time (sec) 2.5 3 3.5 Fig. (5.8) The comparison between the transient response for PI and PID controllers (unit step change). Chapter Five Results and Discussion Fig. (5.5) shows the bode diagram of the closed loop system which implemented to determine the value of ultimate gain (k u ) and ultimate R R period of sustained cycling (P u ) in order to tune the adjusted parameters R R values of both PI and PID modes as in the section (D.3.1) in appendix (D). Figs. (5.6) and (5.7) show the control responses for PI and PID modes for two different criteria. As shown in the figures, it is clear that the overshoot and setting time of Ziegler-Nichols method are less than of the Cohen-Coon method for both PI and PID modes. Fig. (5.8), shows the comparison between two control modes. It is clear that PID mode gave better response that is clear through the lower values of the overshoot and response time. So PID controller is used in this work as a feedback mode of comparison with the other modes of a classical FL, ANN and PID fuzzy controllers. 5.3.2 Fuzzy Logic Controller In this section, classical FL controller is discussed and studied. The control tuning of the FL controller depends on the trial and error to find the scaled factors for each variable. The main difficulty of implementing this FL controller is the number of tuning parameters: the scaled factors for each variable, the membership functions and the rules. The best values of the scaled factors were tuned using simulink program. The simulation model of PHE with classical FL controller is illustrated in Fig. (5.9). Chapter Five Results and Discussion + - step + Mux G Mux Transfer Fcn Valve - Fuzzy Controller with Ruleviewer v G p Transfer Fcn Process Transport Delay Scope -1 Z Integer Delay G m Transfer Fcn measuring Fig. (5.9) Simulation model of PHE with classical FL controller. For the classical FL controller the input variable are error (e) and change of error (de), the output variable is the control action (u). Gaussian membership functions are used for input variables simulations while for the output variable the triangular membership function was used. The universe of discourse of error, delta error and output are [-1, 1], [-0.15, 0.15] and [-1, 1] respectively. The rule base that have been taken proposed by Mamdani fuzzy system. The membership function for error and change of error consist of negative big (NB), negative (N), zero (Z), positive (P) and positive big (PB). Meanwhile, membership function for control action consist of negative big (NB), negative (N), negative small (NS), zero (Z), positive small (PS), positive (P) and positive big (PB). The complete set of classical FL control rules given in table (5.5). The value of membership function for control action is: PB : [0.8 0.9 1] , P : [0.4 0.6 0.7] , PS : [0.1 0.3 0.5] , Z : [-0.2 0 0.2] , NS : [-0.5 -0.3 -0.1] , N : [-0.7 -0.6 -0.4] , NB : [-1 -0.9 -0.8]. Chapter Five Results and Discussion Table (5.5) IF-THEN rule base for classical FL control. de e NB NB N Z P PB PB PB P PS Z N PB P PS Z NS Z P PS Z NS N P PS Z NS N NB PB Z NS N NB NB The table is read in the following way: IF e is NB AND de is NB THEN u is PB. The comparison between the transient response for PID and classical FL controller is shown in Fig. (5.10). Table (5.6) shows the performance criteria for classical FL and PID controllers. As shown in Fig. (5.10) and table (5.6), it's clear that the classical FL controller performs better compared to PID controller in terms of overshoot. But, on comparing the ISE, ITAE, settling time and rise time of both controller, the PID controller performs better because of the trial and error depending of FL controller tuning process. Also there are several reasons that make the PID controller better than classical FL controller [131]: P P • The PID controller is well understood, easy to implement – both in its digital and analog forms – and it is widely used. By contrast, the fuzzy controller requires some knowledge of FL. It also involves building arbitrary membership functions. • The fuzzy controller is generally nonlinear. It does not have a simple equation like the PID, and it is more difficult to analyze mathematically; approximations are required. Chapter Five Results and Discussion • The fuzzy controller has more tuning parameters than the PID controller. Furthermore, it is difficult to trace the data flow during execution, which makes error correction more difficult. The results obtained agreed with the findings of Erenoglu [132]. P PID Controller Classical Fuzzy Controller 2 Outlet Cold Water Temperature (C) P 1.5 1 0.5 0 0 0.5 1 1.5 Time (sec) 2 2.5 3 Fig. (5.10) The comparison between the transient response for PID and classical FL controllers. Table (5.6) Comparison between the performance of classical FL controller and PID controller. Parameters Classical FL controller PID controller ISE 0.4655 0.1236 ITAE 0.2187 0.0298 Overshoot 1.006 2.12 Settling time 2.575 0.73 Rise time 2.715 0.077 Chapter Five Results and Discussion 5.3.3 Artificial Neural Network NARMA-L2 Controller In order to evaluate the effectiveness of the NARMA-L2 control, the controller is implemented and applied to control PHE. NARMA-L2 control is one of the popular neural network architectures that have been implemented as simulink block in MATLAB software version 7.10 which contained in the neural network toolbox blockset. The plant model neural network has one hidden layer of seven neuron which was found as a best neuron number in this work and an output layer of one neuron. The size of hidden layer, the number of delayed inputs and outputs, and the training function are selected in window as shown in Fig. (5.11). This window enable to train and control the plant model. The training function is trainlm and MATLAB simulink used to design the model for the PHE. ANN NARMA-L2 controller simulink block is added to the plant by MATLAB simulink. Fig. (5.12) illustrates the simulation model of PHE with ANN NARMA-L2 controller. Figs. (5.13) and (5.14) show the training and testing of ANN NARMA-L2 controller respectively. These figures contain the input and output of the plant, also it's seen that the NN trained to identify the response of the plant. Fig. (5.15) shows the performance of ANN NARMA-L2 control. The training reached the specified performance of mean square error of (4.654 * 10-9) with 463 epoch. Also the validation and test show the same P P as or close to the training performance. Fig. (5.16) shows the transient response of the PHE with ANN NARMA-L2 controller that shows the outlet cold water temperature (T co ) R R response. This figure shows that the NARMA-L2 gave a good control performance with low values of ISE and ITAE as well as low rise time and settling time as given in table (5.7). Chapter Five Results and Discussion It is clear in Fig. (5.16) and table (5.7) that the ANN NARMA-L2 controller is better than feedback and classical FL controllers because of the good tuning of adjusted parameters values that give smaller ISE, ITAE, and faster to reach the steady state in lower time, lower oscillatory compared with feedback and classical FL controllers but the classical FL controller is better in lower overshoot compared with ANN NARMA-L2 controller. Fig. (5.11) Plant identification window. Chapter Five Results and Discussion Fig. (5.12) Simulation model of PHE with ANN NARMA-L2 controller. Fig. (5.13) Training of ANN NARMA-L2 controller. Chapter Five Results and Discussion Fig. (5.14) Testing of ANN NARMA-L2 controller. Fig. (5.15) The performance of ANN NARMA-L2 control. Chapter Five Results and Discussion ANN NARMA-L2 Controller Outlet Cold Water Temperature (C) 1 0.8 0.6 0.4 0.2 0 0 0.5 1 1.5 Time (sec) 2 2.5 3 Fig. (5.16) Transient response of the PHE with ANN NARMA-L2 controller. Table (5.7) Different performance indices and different parameters of ANN NARMA-L2 controller. Parameters ANN NARMA-L2 controller ISE 0.0601 ITAE 0.0091 Overshoot 1.049 Settling time 0.462 Rise time 0.5 5.3.4 PID Fuzzy Controller The design for classical FL controllers is still considered premature in general, significant progress has been gained recently in the pursuit of Chapter Five Results and Discussion this technology and it remains a difficult task due to the fact that there is insufficient analytical design technique in contrast with the well-developed linear control theories .The FL controller structure can be classified into different types, and the most popular one is PID fuzzy controller. The control tuning of the PID fuzzy controller depends on the trial and error to find the scaled factors for each variable P [133] . The best values P of the scaled factors were tuned using simulink program. The simulation model of PHE with PID fuzzy controller is illustrated in Fig. (5.17). + Gain1 step 1 S G Integrator Gain2 Gain Fuzzy Logic Controller v Transfer Fcn Valve G p Transfer Fcn Process Transport Delay Scope du dt Gain3 G m Transfer Fcn measuring Fig. (5.17) Simulation model of PHE with PID fuzzy controller. The inputs of PID fuzzy control are defined as the proportional gain (K c ), integral gain (K I ) and derivative gain (K D ). The output variable, is R R R R R R called the control action (u). Fuzzy sets are defined for each input and output variable. There are three fuzzy levels (negative (N), zero (Z) and positive (P)). The membership functions for inputs are triangular and the membership function for output variable is linear. By trial and error the proportional gain has a range of [0, 1.25], integral gain has a range of [-2, 2], derivative gain has a range of [0.1, 1.25] and control action has a range of [0, 1.25]. The system is a Sugeno fuzzy system and the rule base of PID fuzzy controller is shown in table (5.8). Chapter Five Results and Discussion Table (5.8) The rule base of PID fuzzy controller. 27 PID fuzzy rules are used for this case because of using three dimensional rule set. For example, one of the rules for PID fuzzy controller: IF K c is N AND K I is N AND K D is N THEN u is P. R R R R R R The transient response of the PHE with PID fuzzy controller is shown in Fig. (5.18) and the performance indices used of PID fuzzy controller are the ISE and ITAE as well as the performance of the PID fuzzy controller of the step response of the system are given in table (5.9). As shown in Fig. (5.18) and table (5.9), it's clear that the PID fuzzy controller is improved over to other controllers used in this work. Chapter Five Results and Discussion PID Fuzzy Controller Outlet Cold Water Temperature (C) 1 0.8 0.6 0.4 0.2 0 0 1 0.5 2 1.5 Time (sec) 2.5 3 Fig. (5.18) Transient response of the PHE with PID fuzzy controller. Table (5.9) Different performance indices and different parameters in PID fuzzy controller. Parameters PID fuzzy controller ISE 0.0547 ITAE 0.0031 Overshoot 1 Settling time 0.432 Rise time 0.599 Chapter Five Results and Discussion 5.3.5 Comparison Among PID, Artificial Neural Network and PID Fuzzy Controllers This section shows a comparison among different control strategies of the transient response for the PHE with PID, ANN and PID fuzzy controllers as shown in Fig. (5.19). To evaluate the performance of different controllers parameters of the step response of the system have been considered. In all the three controllers the performance indices of different controllers are the ISE and ITAE as well as the parameters are evaluated and comparative studies of their performance are tabulated in the table (5.10). PID Fuzzy Controller ANN Controller PID Controller Outlet Cold Water Temperature (C) 2 1.5 1 0.5 0 0 0.5 1 1.5 Time (sec) 2 2.5 3 Fig. (5.19) The comparison among the transient response for PID, ANN and PID fuzzy controllers. Chapter Five Results and Discussion Table (5.10) Comparison of different performance indices and different parameters in controllers. PID ANN PID fuzzy controller controller controller ISE 0.1236 0.0601 0.0547 ITAE 0.0298 0.0091 0.0031 Overshoot 2.12 1.049 1 Settling time 0.73 0.462 0.432 Rise time 0.077 0.5 0.599 Parameters From Fig. (5.19) and table (5.10), the simulation results clearly show that the PID fuzzy controller gives better control of temperature rather than PID controller and ANN controller. It has been seen that more accurate results were obtained using ANN controller over PID controller, further better results obtained by using PID fuzzy controller. From the above observations it is clear that the PID controller produces high values of overshoot and settling time. To compensate this kind of high values, ANN controller has been implemented. By implementing this method the system overshoot and settling time were reduced. For further reduction requirements, the PID fuzzy controller was suggested. ISE and ITAE of PID fuzzy controller show lower values compared to other modes which indicates the robust control of this controller. Although PID fuzzy mode gave better performance, but the high value of the rise time shows one of its disadvantages. The reason for that high value is the significant time investment needed to correctly tune membership functions and adjust rules to obtain a good solution. The more rules suggested, the increasing difficulty obtained. The results required more system memory and processing time [133]. P P Chapter Five Results and Discussion The results showed that the PID fuzzy controller is slightly better than ANN controller. From these observations it is clear that PID fuzzy controller is a much better option for control rather than PID and ANN controller because PID fuzzy controller combines the advantages of a fuzzy logic controller and a PID mode. Also PID fuzzy controller is decreasing the number of rules, decomposition of multivariable control rules into three sets of one dimensional rules for each input variable, simplified the evolution of the rule base, conventional control and easy connection between fuzzy parameters and operation of the controller and membership functions are simple triangular with fuzzy logic rules. A major problem with neural nets is the “Black Box” nature, or rather, the relationships of the weight changes with the input-output behavior during training and use of trained system to generate correct outputs using the weights. Chapter Six Conclusions and Future Work 6.1 Conclusions Based on this study of dynamics and control of a PHE, the following conclusions can be derived: 1. The experimental heat transfer measurements of the PHE show that the overall heat transfer coefficient (U) is related to the hot water flow rate (m h ) by a correlation having the form: R R U = 11045 m 0.7158 h ……….. (6.1) 2. The PHE model is found dynamically as a first order lead and second order overdamped lag while the experimental PHE represented dynamically as first order with a negligible dead time value. 3. The assumptions used to establish the mathematical model of present process give good agreement when the theoretical and experimental results are compared with each other. 4. The response of 24 plates when justifying the design fitness of the PHE prove that the design is accurate and there is no losses in the energy input. 5. An important step in implementing the mathematical model of the PHE is the selection of the controller parameters. Different methods were used in the computer simulation to optimize the numerical value of controller parameters, the ITAE gives a good comparison and clearance of the error. 6. PID feedback controller is better than PI feedback controller because it gives smaller ITAE, ISE, overshoot, settling time and rise time values. 7. PID controller performs better when it is compared to classical fuzzy logic controller because of the trial and error depending of fuzzy logic controller tuning process. Chapter Six Conclusions and Future Work 8. Artificial neural network controller is better than feedback and classical fuzzy logic controllers because the artificial neural network controller learns system and it has got generalization capabilities. 9. The PID fuzzy controller gives a much better control performance of temperature rather than PID controller and artificial neural network controller because PID fuzzy controller combines the advantages of fuzzy logic controller and a PID controller. 6.2 Future Work The following suggestions for future work can be considered: 1. The same procedure of this work is useful for another type of heat exchanger with different specifications or using the same procedure for other controlled and manipulated variables. 2. Adding other control strategies like cascade control, adaptive control, neuro-fuzzy control and genetic algorithms control. 3. Application of on-line control is recommended. References 1. Bayazit, S., Bicer, K. H., Kulali, G., Müminoglu, M., and Torres, J. J. J., "Automatic Control of a Heat Exchanger with Changing Operation Conditions", http://www.mathematik.uniU dortmund.de/papers/BayazitBicerKulaliMueminogluTorres2008.pd f, 2008. U 2. Diaz, G., Sen, M., Yang, K. T., and McClain, R. L., "Adaptive Neurocontrol of Heat Exchangers", Transactions of the ASME, Journal of Heat Transfer, Vol.123, PP.556-562, June, 2001. 3. Patel, A. M., "Dynamic Mathematical Model of a Heat Exchanger", PhD Thesis, Chemical Engineering, University of Tennessee at Chattanooga, March, 2003. 4. Hu, Q., So, A. T. P., Tse, W. L., and Ren, Q., "Development of ANN-Based Models to Predict The Static Response and Dynamic Response of a Heat Exchanger in a Real MVAC System", International Conference on Control and Synchronization of Dynamical Systems, China, Journal of Physics, Vol.23, PP.110121, 2005. 5. "Heat Exchanger", http://www.answers.com/topic/heatU U0T U0T exchanger, November 2010. U 6. Haile, G. T., "Heat Transfer in Plate Heat Exchangers", M.Sc., Thesis, Chemical and Process Engineering, Lappeenrata University of Technology, June 2009. 7. Fernandes, C. S., Dias, R. P., and Maia, J. M., "New Plates for Different Types of Plate Heat Exchangers", Recent Patents on Mechanical Engineering, Vol.1, No.3, PP.198-205, 2008. 8. "APV Plate Heat Exchangers", APV ANSPX BRAND, http://www.apv.com , 2010. 0TU U0T References 9. "Plate Heat Exchangers", WCR Sweden Specialist in Plate Heat Exchangers, http://wcrbenelux.nl/site/content/view/12/13/lang ,en, 0TU U0T U 2009. 10. Narataruksa, P., Triratana, P., Suppamassadu, K. P., Heggs, P. J., and Tia, S., "Dynamic Simulation of Plate and Frame Heat Exchanger Undergoing Food Fouling: Coconut Milk Fouling Case Study", www.scienceasia.org, Jan 2008. 0TU U0T 11. Gut, J. A. W., and Pinto, J. M., "Modeling of Plate Heat Exchangers with Generalized Configurations", International Journal of Heat and Mass Transfer, Vol.46, PP.2571-2585, 2003. 12. Thirumarimurugan, M., and Kannadasan, T., "Simulation Studies on Plate Type Heat Exchanger Using ANN", International Journal of Chem. Tech. Research, Vol.1, No.2, PP.349-354, AprilJune, 2009. 13. Dwivedi, A. k., and Das, S. K., "Dynamics of Plate Heat Exchangers Subject to Flow Variations", International Journal of Heat and Mass Transfer, Vol.50, PP.2733-2743, 2007. 14. Williams, D., "Plate Heat Exchangers", Canada Composting Inc. (CCI), State of the Art Engineering Report-MME 499, October, 2002. 15. Longo, G. A., and Gasparella, A., "Refrigerant R134a Vaporization Heat Transfer and Pressure Drop Inside A Small Brazed Plate Heat Exchanger", International Journal of Refrigeration, Vol.30, PP.821-830, 2007. 16. Gut, J. A. W., Fernandes, R., Pinto, J. M., and Tadini, C. C., "Thermal Model Validation of Plate Heat Exchangers with Generalized Configurations", Chemical Engineering Science , Vol.59, PP.4591-4600 , 2004. References 17. Galeazzo, F. C. C., Miura, R. Y., Gut, J. A. W. and Tadini, C. C., "Experimantal and Numerical Heat Transfer in a Plate Heat Exchanger", Chemical Engineering Science, Vol.61, PP.71337138, 2006. 18. Alfa Laval, Product Catalogue, Sweden, http://www.alfalaval.com. 0TU U0T 19. Seborg, D. E., Edgar, T. F., and Mellichamp, D. A., "Process Dynamics and Control", 2nd Edition, John Wiley and Sons, Inc., P P 2004. 20. Ali, E. M., "Process Control in The Chemical Industries", http://ksu.edu.sa/Pages/default.aspx, 2002. 0TU U0T 21. Kaya, I., Tan, N., and Atherton, D. P., "Improved Cascade Control Structure and Controller Design", 44th P P IEEE Conference on Decision and Control and the European Control Conference, Seville, Spain, December, 2005. 22. You, K., "Adaptive Control", In-Tech, www.intechopen.com, U U January, 2009. 23. Grigorie, T. L., "Fuzzy Controllers, Theory and Applications", In-Tech, www.intechopen.com, February, 2011. U U 24. Luebbers, P. G., "Process-Industrial Instruments and Control Handbook", Iocalhost, http://localhost.com, 2004. 0TU U0T 25. Mastacan, L., Olah, I., and Dosoftie, C. C., "District Heating Substations Water Temperature Control Based on Soft Computing Technology", 6th International Conference on P P Electromechanical and Power System, Chisinau, Rep.Moldova, October, 2007. 26. Roffel, B., and Betlem, B., "Process Dynamics and Control", John Wiley and Sons, Ltd, 2006. 27. Scariot, M. R., Berto, M. I., and Silveira, V., "Simulation of The Temperature Profile of a Pectin Solution in a Plate Heat References Exchanger: a Non-Linear System Approach for Control", Thermal Engineering, Vol.4, No.1, PP.30-34, June, 2005. 28. Alwan, G. M., "Study of Dynamics and Control of a Plate Heat Exchanger", Msc Thesis, Chemical Engineering, University of Baghdad, November, 1982. 29. Baker, N. S., "Dynamic Characteristics of Plate Heat Exchanger", Msc Thesis, Chemical Engineering, University of Technology, December, 1983. 30. Khan, A. R., Baker, N. S., and Wardle, A. P., "The Dynamic Characteristics of a Counter Current Plate Heat Exchanger", Int. J. Heat Mass Transfer, Vol.31, No.6, PP.1269-1278, 1988. 31. AL-Zobai, K. M. M., "Computer Control on Plate Heat Exchanger", PhD Thesis, Chemical Engineering, University of Baghdad, 2004. 32. Kapustenko, P., Dobromyslova, O., Dobromyslov, O., Perevertaylenko, O., Arsenyeva, O., Ilyunin, O., and Shabanov, E., "Control of Plate Heat Exchanger Outlet Temperature Using Butterfly Valve and Parametric Model Predictive Control Technique", Chemical Engineering Transactions, Vol.18, 2009. 33. Mandavgane, S. A., and Pandharipande, S. L., "Application of Optimum ANN Architecture for Heat Exchanger Modeling", Indian Journal of Chemical Technology, Vol.13, PP.634-639, November, 2006. 34. Vasickaninova, A., Bakosova, M., Meszaros, A., and Klemes, J., "Neural Network Predictive Control of a Heat Exchanger", Chemical Engineering Transactions, Vol.21, 2010. 35. Prasanth, B. V., and Kumar, D. S. V. J., "New Control Strategy for Load Frequency Problem of a Single Area Power System References Using Fuzzy Logic Control", Journal of Theoretical and Applied Information Technology, 2008. 36. Anand, B., and Jeyakumar, A. E., "Load Frequency Control with Fuzzy Logic Controller Considering Non-Linearities and Boiler Dynamics", ICGST-ACSE Journal, Vol.8, Issue III, January, 2009. 37. Xue, D., Chen, Y. Q., and Atherton, D. P., "Linear Feedback Control", The Society for Industrial and Applied Mathematics, 2007. 38. Imal, E., "CDM Based Controller Design for Nonlinear Heat Exchanger Process", Turkey Journal Electrical Engineering and Computer Science, Vol.17, No.2, 2009. 39. Henriques, J., Gil, P., Cardoso, A., and Dourado, A., "Scheduling of PID Controllers by Means of a Neural Network with Application to a Solar Power Plant", The World's Largest Professional Association for the Advancement of Technology, IEEE, 2002. 40. Delnero, C., "Neural Networks and PI Control Using Steady State Prediction Applied to a Heating Coil", Msc Thesis, Mechanical Engineering, Colorado State University, October, 2000. 41. Fink, A., Nelles, O., and Isermann, R., "Nonlinear Internal Model Control for Miso Systems Based on Local Linear NeuroFuzzy Models", IFAC, 15th Triennial World Congress Barcelona, P P Spain, 2002. 42. Maidi, A., Diaf, M., and Corriou, J. P., "Optimal Linear PI Fuzzy Controller Design of a Heat Exchanger", Chemical Engineering and Processing, Vol.47, PP.938-945, 2008. References 43. Dale, S., Gabor, G., and Bara, A., "Reduced Complexity Interpolative Control System for a Geothermal Heat Exchanger", World Geothermal Congress, Bali, Indonesia, April, 2010. 44. Diaz, G. C., "Simulation and Control of Heat Exchangers Using Artificial Neural Networks", PhD Thesis, Aerospace and Mechanical Engineering, University of Notre Dame, March, 2000. 45. Rouhani, H., Kharaajoo, M. J., Araabi, B. N., and Lucas, C., "Intelligent Control of Electrically Heated Micro Heat Exchanger with Locally Linear Neurofuzzy Identifier and Emotional Based Learning Controller", The World Scientific and Engineering Academy and Society (WSEAS), Conferences, Austria, 2004. 46. Varshney, K., and Panigrahi, P. K., "Artificial Neural Network Control of a Heat Exchanger in a Closed Flow Air Circuit", Applied Soft Computing, Vol.5, PP.441-465, 2005. 47. Qu, Y., Xu, L., Fang, X., Wang, J., and Gu, S., "A New Approach to Heat Exchanger Control Based on Model Control", International Journal of Information and Systems Sciences, Vol.2, No.1, PP.31-41, 2006. 48. Jan, M., and Jaroslava, K., "Control of Laboratory Pilot Plant with Dead Time", Instruments and Control, ASR, Babiuch, Ostrava, 2009. 49. Mulyana, T., Than, M. N. M., and Hanafi, D., "A Discrete Time Model of Boiler Drum and Heat Exchanger QAD Model BDT 921", International Conference on Instrumentation, Control and Automation, Bandung, Indonesia, October, 2009. 50. Homod, R. Z., Sahari, K. S. M., Mohamed, H. A. F., and Nagi, F., "Hybrid PID-Cascade Control for HVAC System", References International Journal of Systems Control, Vol.1, Iss.4, PP.170-175, 2010. 51. Khare, Y. B., and Singh, Y., "PID Control of Heat Exchanger System", International Journal of Computer Applications, Vol.8, No.6, October, 2010. 52. Padhee, S., and Singh, Y., "A Comparative Analysis of Various Control Strategies Implemented on Heat Exchanger System: A Case Study", World Congress on Engineering (WCE), London, U.K., Vol.II, 2010. 53. Rajasekaran, R., and Kannadasan, T., "A Simplified Predictive Control for a Shell and Tube Heat Exchanger", International Journal of Engineering Science and Technology, Vol.2, No.12, PP.7245-7251, 2010. 54. Vasickaninova, A., and Bakosova, M., "Locally Optimal Fuzzy Control of a Heat Exchanger", The World Scientific and Engineering Academy and Society (WSEAS), Transactions on Systems, Vol.9, Issue 9, September, 2010. 55. Riverol, C., and Naplitano, V., "Use of Neural Networks as a Tuning Method for an Adaptive PID Application in a Heat Exchanger", Transactions Institution of Chemical Engineers, Vol.78, Part A, November, 2000. 56. Diaz, G., Sen, M., Yang, K. T., and McClain, R. L., "Dynamic Prediction and Control of Heat Exchangers Using Artificial Neural Networks", International Journal of Heat and Mass Transfer, Vol.44, PP.1671-1679, 2001. 57. Berto, M. I., and JR, V. S., "Configuration of PID / Feed Back and PID / Feed Back / Feed Forward Controllers in Temperature Control of a HTST Heat Exchanger", 2th P Mercosur Congress on Chemical Engineering and 4th Mercosur P P P References Congress on Process Systems Engineering, Sao Paulo, Brazil, 2004. 58. Juneja, P. K., Ray, A. K., and Mitra, R., "Fuzzy Control and Neural Network Control of Limekiln Process", International Journal of Electronics Engineering, Vol.2, No.2, PP.305-306, 2010. 59. Xie, G. N., Wang, Q. W., Zeng, M., and Lou, L. Q., "Heat Transfer Analysis for Shell-and-Tube Heat Exchangers with Experimental Data by Artificial Neural Networks Approach", Applied Thermal Engineering, Vol.27, PP.1096-1104, 2007. 60. Cam, E., and Kocaarslan, I., "Load-Frequency Control in Two Area Power System", Teknoloji, Vol.7, Issue 2, PP.197-203, 2004. 61. Piegat, A., "What is Not Clear in Fuzzy Control Systems?", International Journal Applied Mathematic Computer Science, Vol.16, No.1, PP.37-49, 2006. 62. Malhotra, R., Singh, N., and Singh, Y., "An Efficient Fuzzy-GA Flow Control of Turbine Compressor System: A Process Control Case Study", International Journal of Advancements in Computing Technology, Vol.2, No.4, October, 2010. 63. Skrjance, I., Blazic, S., and Matko, D., "Model-Reference Fuzzy Adaptive Control as a Framework for Nonlinear System Control", Journal of Intelligent and Robotic Systems, Vol.36, PP.331-347, 2003. 64. Dale, S., Bara, A., and Gabor, G., "Interpolative Control Structure Design for a Heat Exchanger in a Geothermal Power Plant", Journal of Computer Science and Control Systems, Vol.1, Issue.1, PP.131-134, 2008. 65. Auttawaitkul, Y., Monyakul, V., and Therdyothin, A., "A new Model of Variable Air Volume Diffuser for Thermal Comfort References Based on Fuzzy Logic Control", ECTI International Conference, Thailand, 2007. 66. Kolek, L., and Harmati, I., "Model Based Simulation and Fuzzy Control of Solar Heating Systems", 13th IEEE IFAC International P P Conference on Methods and Models in Automation and Robotics, Szczecin, Poland, August, 2007. 67. Mazinan, A. H., and Sadati, N., "Multiple Modeling and Fuzzy Predictive Control of a Tubular Heat Exchanger System", The World Scientific and Engineering Academy and Society (WSEAS), Transactions on Systems and Control, Vol.3, Issue 4, April, 2008. 68. Skrjanc, I., and Matko, D., "Predictive Functional Control Based on Fuzzy Model for Heat-Exchanger Pilot Plant", IEEE Transactions on Fuzzy Systems, Vol.8, No.6, December, 2000. 69. Chen, Y., Zhang, J., Zhang, B., and Gao, F., "Development of Monitoring Control and Fuzzy Control Test of Finned-Tube Heat-Exchanger Test-Board", Sixth International Conference for Enhanced Building Operations, Shenzhen, China, Vol.VI-6-1, November, 2006. 70. Habbi, H., Kinnaert, M., and Zelmat, M., "A Complete Procedure for Leak Detection and Diagnosis in a Complex Heat Exchanger Using Data-Driven Fuzzy Models", ISA Transactions, Vol.48, PP.354-361, 2009. 71. Agashe, S., Ghatol, A., and Agashe, S., "Automation of Heat Exchanger Using Neural Network", World Academy of Science, Engineering and Technology, Vol.15, 2006. 72. Renotte, C., Wouwer, A. V., and Remy, M., "Neural Modeling and Control of a Heat Exchanger Based on SPSA Techniques", American Control Conference, Chicago, Illinois, PP.3299-3303, June, 2000. References 73. Babu, B. V., and Shailesh, M., "Adaptive Networks for Fault Diagnosis and Process Control", Birla Institute of Technology and Science, India, www.bits-pilani.ac.in, 2002. U U 74. Kharaajoo, M. J., "Neural Network Control of a Heat Exchanger Pilot Plant", The World Scientific and Engineering Academy and Society (WSEAS), 2003. 75. Riverol, C., and Napolitano, V., "Estimation of Fouling in a Plate Heat Exchanger Through The Application of Neural Networks", Journal of Chemical Technology and Biotechnology, Vol.80, PP.594-600, 2005. 76. Ramasamy, M., Zabiri, H., Ha, N. T. T., and Ramli, N. M., "Heat Exchanger Performance Prediction Modeling Using NARXType Neural Networks", The World Scientific and Engineering Academy and Society (WSEAS), International Conference on Waste, Management, Water Pollution, Air Pollution, Indoor Climate, Arcachon, France, October, 2007. 77. Dudzik, S., "Calculation of The Heat Power Consumption in The Heat Exchanger Using Artificial Neural Network", 9th P P International Conference on Quantitative Infrared Thermography, Krakow, Poland, July, 2008. 78. Fadare, D. A., and Fatona, A. S., "Artificial Neural Network Modeling of Heat Transfer in a Staggered Cross-Flow TubeType Heat Exchanger", Pacific Journal of Science and Technology, Vol.9, No.2, November, 2008. 79. Ramasamy, M., Shahid, A. and Zabiri, H., "Drift Analysis on Neural Network Model of Heat Exchanger Fouling", Journal of Engineering Science and Technology, Vol.3, No.1, PP.40-47, 2008. 80. Bonala, S. Y., "A Study on Neural Network Based System Identification with Application to Heating Ventilating and Air References Conditioning (HVAC) System", Msc Thesis, Electrical Engineering, National Institute of Technology, Rourkela, May, 2009. 81. Martinez, J. C. T., Menendez, R. M., and Castanon, L. E. G., "Fault Diagnosis in a Heat Exchanger Using Process History Based-Methods", 20th European Symposium on Computer Aided P P Process Engineering-ESCAPE20, Monterrey, Mexico, 2010. 82. Kharaajoo, M. J., Araabi, B. N., "Neural Network Based Predictive Control of a Heat Exchanger Nonlinear Process", Journal of Electrical and Electronics Engineering, Vol.4, No.2, PP.1219-1226, 2004. 83. Farahani, S. S. A., Nekoui, M. A. and Goharrizi, A. Y., "Predictive Control of a Heat Exchanger Based on Local Fuzzy Models and Neural Networks", Journal of Electrical and Electronics Engineering, Vol.6, No.2, PP.129-138, 2006. 84. Biyanto, T. R., Ramasamy, M., and Zabiri, H., "Modeling Heat Exchanger Using Neural Networks", International Conference on Intelligent and Advanced Systems, Perak, Malaysia, 2007. 85. Selbas, R., Sencan, A., and Kilic, B., "Alternative Approach in Thermal Analysis of Plate Heat Exchanger", Heat Mass Transfer, Vol.45, PP.323-329, 2009. 86. APV Company Ltd. England type (JHE) serial No. (1062), 1972. 87. Peter, R. N. C., "Practical Temperature Measurement", Butterworth-Heinemann, 2001. 88. Stephanopoulos, Introduction G., to "Chemical Theory and Process Practice", Control An Prentice-Hall International Inc., London, 1984. 89. Luyben, M. L., and Luyben, W. L., "Essential of Process Control", McGraw-Hill Companies Inc., 1997. References 90. Nagaraj, B. ,Muthusamy, P., Murali, B., Shahulhammed, M., and Murugananth, N., "Optimum Tuning Algorithms for PID Controller A Soft Compuing Approach", IPPTA Journal, Vol.22, No.2, June, 2010. 91. Cohen, G. H., and Coon, G. A., "Theoretical Considerations of Retarded Control", Trans. ASME, Vol.75, PP.827, 1953. 92. Ziegler, J. G., and Nichols, N. B., "Optimum Settings for Automatic Controllers", Trans. ASME, Vol.64, PP.759, 1942. 93. Chrysostomou, C., "Fuzzy Logic Based AQM Congestion Control in TCP/IP Networks", PhD Thesis, Computer Science, University of Cyprus, September, 2006. 94. Srivastava, A. K., Kamalasadan, S., and Hande, A., "Comparative Performance of Improved Shrinking Span Fuzzy Logic Controller", The World's Largest Professional Association for the Advancement of Technology, IEEE, 2006. 95. Mathur, H. D., and Manjunath, H. V., "Frequency Stabilization Using Fuzzy Logic Based Controller for Multi-Area Power System", The South Pacific Journal of Natural Science, Vol.4, PP.22-30, 2007. 96. Pundaleek, B. H., Rathi, M. G., and Vijay, K. M. G., "Speed Control of Induction Motor: Fuzzy Logic Controller v/s PI Controller", International Journal of Computer Science and Network Security, Vol.10, No.10, October, 2010. 97. Monga, H., Kaur, G., Kaur, A., and Soni, K., "Fuzzy Logic Controller for Analysis of AGC", International Journal of Advanced Engineering and Applications, January, 2010. 98. Cirstea, M. N., Dinu, A., Khor, J. G., and Cormick, M. M., "Neural and Fuzzy Logic Control of Drives and Power Systems", Newnes, www.newnespress.com, 2002. U U References 99. Zhang, H., and Liu, D., "Fuzzy Modeling and Fuzzy Control", Birkhauser Boston, www.birkhauser.com, 2006. U U 100. Corcau, J. I., and Stoenescu, E., "Fuzzy Logic Controller as a Power System Stabilizer", International Journal of Circuits, Systems and Signal Processing, Vol.1, Issue 3, 2007. 101. Gayakwad, R., "Optimized Fuzzy Logic for Motion Control", Acta Polytechnica Hungarica, Vol.7, No.5, 2010. 102. Mortazavi, S. S., Razaz, M., and Khavari, E., "Power Quality Improvement Using a Fuzzy Logic Control of a Series Active Filter", Journal of Theoretical and Applied Information Technology, 2010. 103. Sakthivel, G., Anandhi, T. S., and Natarajan, S. P., "Real Time Implementation of a Fuzzy Logic Controller on FPGA Using VHDL for DC Motor Speed Control", International Journal of Engineering Science and Technology, Vol.2, No.9, PP.4511-4519, 2010. 104. Wahyudi, and Jalani, J., "Design and Implementation of Fuzzy Logic Controller for Intelligent Gantry Crane System", 2nd P P International Conference of Mechatronics, Kuala Lumpur, Malaysia, May, 2005. 105. Peri, V. M., "Fuzzy Logic Controller for an Autonomous Mobile Robot", Msc Thesis, Electrical Engineering, Cleveland State University, May, 2005. 106. Jha, P., "Novel Artificial Neural Network Application for Prediction of Inverse Kinematics of Manipulator", Msc Thesis, Mechanical Engineering, National Institute of Technology, 2009. 107. Haghizadeh, A., Shui, I. T., and Goudarzi, E., "Estimation of Yield Sediment Using Artificial Neural Network at Basin References Scale", Australian Journal of Basic and Applied Sciences, Vol.4, No.7, PP.1668-1675, 2010. 108. Dase, R. K., and Pawar, D. D., "Application of Artificial Neural Network for Stock Market Predictions: A Review of Literature", International Journal of Machine Intelligence, Vol.2, Issue 2, PP.14-17, 2010. 109. Basu, J. K., Bhattacharyya, D., and Kim, T. H., "Use of Artificial Neural Network in Pattern Recognition", International Journal of Software Engineering and Its Applications, Vol.4, No.2, April, 2010. 110. Almhdi, K. M., Valigi, P., Gulbinas, V., Westphal, R., and Reuter, R., "Classification with Artificial Neural Networks and Support Vector Machines : Application to Oil Fluorescence Spectra", European Association of Remote Sensing Laboratories (EARSEL), www.eproceedings.org, 2007. U U 111. Jha, G. K., "Artificial Neural Networks", Indian Agricultural Research Institute, PUSA, New Delhi, 2004. 112. Paulraj, M. P., Yaacob, S., and Andrew, A. M., "Vehicle Noise Comfort Level Indication Using Artificial Neural Network", 3rd P Regional Conference on Noise, Vibration and Comfort (NVC), Putrajaya, Malaysia, June, 2010. 113. Jena, G., and Baliarsingh, R., "An Introductory Course on Artificial Neural Network and Fuzzy Logic, The World's Largest Professional Association for the Advancement of Technology, IEEE, http://www.drgjena.co.cc, 2009. U U 114. Abraham, A., "Handbook of Measuring System Design", John Wiely and Sons, Ltd., 2005. 115. Krogh, A., "What are Artificial Neural Networks?", Nature Biotechnology, Vol.26, No.2, February, 2008. P References 116. Messa, V., "Neural Networks for Diagnosis of Infarct and Ischemia", Msc Thesis, IT-University of Goteborg, May, 2006. 117. Lee, Y., "A Neural Network Face Detector Design Using BitWidth Reduced FPU in FPGA", Msc Thesis, Electrical and Computer Engineering, University of Saskatchewan Saskatoon, January, 2007. 118. Kamruzzaman, J., Begg, R. K., and Sarker, R. A., "Artificial Neural Networks in Finance and Manufacturing", Idea Group Inc., http://www.idea-group.com, 2006. U U 119. Du, K. L., and Swamy, M. N. S., "Neural Networks in a Soft Computing Framework", Springer-Verlag London Limited, 2006. 120. Bomma, P., "Computer-Aided Diagnosis Tool for the Detection of Cancerous Nodules in X-Ray Images", Msc Thesis, Electrical and Computer Engineering, Louisiana State University, May, 2005. 121. Georgios, T., Dimitrios, B., Dimitrios, P., and Luis, L. J., "Comparative Control of a Nonlinear First Order Velocity System by a Neural Network NARMA-L2 Method", Elektronika IR Elektrotechnika, Vol.55, No.6, 2004. 122. Oliveira, J. C. P. D., and Bauchspiess, A., "Neural Control of a 4rd Order Level Process", ABCM Symposium Series in Mechatronics, Vol.3, PP.197-206, 2008. 123. Mokri, S. S., Husain, H., Martono, W., and Shafie, A., "Real Time Implementation of NARMA-L2 Control of a Single Link Mainipulator", American Journal of Applied Sciences, Vol.5, No.12, PP.1642-1649, 2008. 124. Atasoy, I., Yuceer, M., and Berber, R., "Molecular Weight Control in Acrylonitrile Polymerization with Neural Network Based Controllers", 16th European Symposium on Computer P P References Aided Process Engineering and 9th International Symposium on P P Process Systems Engineering, Ankara, Turkey, 2006. 125. Beale, M. H., Hagan, M. T., and Demuth, H. B., "Neural Network ToolboxTM 7", The Math Works, Inc., 2010. P P 126. Araghi, L. F., and Koorayme, M. H., "Neural Network Controller for Two Links-Robotic Manipulator Control with Different Load", International Multi Conference of Engineers and Computer Scientists, Hong Kong, Vol II, March, 2009. 127. Chapman, S. J., "Matlab Programming for Engineers", Bookwave Companion Series, Second Edition, 2004. 128. Ayub, Z. H., "Plate Heat Exchanger Literature Survey and New Heat Transfer and Pressure Drop Correlations for Refrigerant Evaporators", Heat Transfer Engineering, Vol.24, No.5, PP.3-16, 2003. 129. Durmus, A., Benli, H., Kurtbas, I., And Gul, H., "Investigation of Heat Transfer and Pressure Drop in Plate Heat Exchangers Having Different Surface Profiles", International Journal of Heat and Mass Transfer, Vol.52, PP.1451-1457, 2009. 130. Chapra, S. C., and Canale, R. P., "Numerical Methods for Engineers with Personal Computer Applications", McGraw-Hill, Inc., 1985. 131. Jantzen, J., "Foundations of Fuzzy Control", John Wiley and Sons, Ltd, 2007. 132. Erenoglu, I., Eksin, I., Yesil, E., and Guzelkaya, M., "An Intelligent Hybrid Fuzzy PID Controller", 20th European P P Conference on Modelling and Simulation Wolfgang Borutzky, Alessandra Orsoni, Richard Zobel, 2006. References 133. Jain, L. C., and Martin, N. M., "Fusion of Neural Networks, Fuzzy Systems and Genetic Algorithms: Applications", www.itknowledge.com, 1998. 0TU U0T Industrial Appendix A Calibration Curves of Thermocouples 4.5 4 3.5 mVolt 3 2.5 mV experimental Linear (fit) 2 1.5 1 0.5 0 0 50 100 150 Temperature (°C) Fig. (A.1) Calibration curve of the thermocouple. volume flowrate (lit/min) 12 10 8 Series1 6 Linear (Series1) 4 2 0 0 2 4 6 8 10 Rotameter reading (lit/min) Fig. (A.2) Calibration curve of cold water rotameter. volume flowrate (lit/min) 9 8 7 6 5 Series1 4 Linear (Series1) 3 2 1 0 0 2 4 6 8 Rotameter reading (lit/min) Fig. (A.3) Calibration curve of hot water rotameter. Appendix B System and Operating Conditions Table (B.1) System and operating conditions. Cold water flow rate mc 0.0414 (kg/sec) Hot water flow rate mh 0.0497 (kg/sec) Temperature of initial cold water T ci 20 (oC) Temperature of initial hot water T hi 50 (oC) Specific heat capacity Cp 4174 (J/kg.oC) Water density Ro 993 (kg/m3) Length of plate heat exchanger L 0.58 (m) Width of plate heat exchanger E 0.07 (m) Thickness of one plate heat exchanger S 0.001 (m) Length of fluid in plate heat exchanger L_ fluid 0.50 (m) Width of fluid in plate heat exchanger E_ fluid 0.065 (m) Thickness of fluid in plate heat exchanger S_ fluid 0.046 (m) B.1 Appendix C Controller Tuning Methods C.1) Cohen-Coon Method [91] Cohen-Coon used process reaction curve, that it is a response of the process to a step change in the manipulated variable. Cohen and Coon observed that the response of most processing units to a step change in input variable can be adequately approximated by the response of first order system with dead time, and the transfer function is: G PRC (s) − K ets τ s +1 = ……….. (C.1) d The values of K, τ and t d are calculated from the process reaction curve which is shown in Fig. (C.1). A tangent is drawn to the curve at the point of maximum rate or ascent, and then t d is the intercept of this tangent with x-axis, and is defined as the time elapsed until the system responds. K = B Output (at steady state) = A Input (at steady state) B Output (at steady state) = S Slope τ = ……….. (C.2) ……….. (C.3) 1) For Proportional controller: K C = 1 τ + t 1 K t 3τ d d ……….. (C.4) 2) For Proportional-Integral controller: K C = 1 τ t 0.9 + 12τ K t d d ……….. (C.5) 30 + 3t τ τ = t 9 + 20t τ ……….. (C.6) d I d d 3) For Proportional-Integral-Derivative controller: K C = 1 τ 4 + t K t 3 4τ d d ……….. (C.7) 32 + 6t τ τ = t 13 + 8t τ ……….. (C.8) d I d d C.1 Appendix C τ = td D Controller Tuning Methods 4 11 + 2t τ ……….. (C.9) d Actual response ym ym B B Approximate response S Slop = S t td t td (b) (a) Fig. (C.1) (a) Temperature curve for Cohen-Coon tuning. (b) Temperature curve approximation with a first order dead-time system. C.2) Ziegler-Nichols Method [92] Ziegler-Nichols used bode diagram of two graphs: one is a plot of the logarithm of the magnitude of sinusoidal transfer function; the other is a plot of phase angle; both are plotted against the frequency on a logarithm scale as shown in Fig. (C.2). Gain margin (GM) and crossover frequency ( ω ) can be found from two plots therefore, the ultimate gain and period of oscillation are calculated from following: K = 20 log(GM ) ……….. (C.10) u P u = 2×3.1428 ……….. (C.11) ω C.2 Appendix C Controller Tuning Methods 1) For Proportional controller: K K 2 = C ……….. (C.12) u 2) For Proportional-Integral controller: K 2.2 ……….. (C.13) τ = 1P.2 ……….. (C.14) K C = u u I 3) For Proportional-Integral-Derivative controller: K K 1.7 ……….. (C.15) = P 2 ……….. (C.16) = P 8 C τ I τ D = u u ……….. (C.17) u 1.0 Gain margin A.R M 0o Ø(I) Phase margin Ø -180o ω ωCO Figure (C.2): Definition of gain and phase margins. C.3 Appendix D MATLAB Program D.1 Introduction This appendix discusses the computer program developed for the dynamic model and control for both open loop and closed loop of system. All programs were developed using MATLAB program version 7.10. The use of friendly, easy to use interfaces for these programs were efficient in the implementation of the model procedure that gave dynamic results obtained from the computer programs. Each program was executed and the results were checked to meet the model requirements, then–if necessary– the design data was modified to meet the requirements of the model. Table (D.1) lists some functions and commands and their description that were used in computer simulation for dynamic behavior and controller design [127]. P P D.1 Appendix D MATLAB Program Table (D.1) Summary functions in MATLAB program. Function Name Function Description Pade Computes the an nth-order approximation to a time delay Series Computes a series system connection Tf Creates a transfer function model object Step Calculates a unit step response of a system Figure Creates new figure window Plot Generates a linear plot Xlabel Add the label to the x-axis of the current graph Ylabel Add the label to the y-axis of the current graph Axis Specific the manual axis scaling on plot Title Add a title to the current graph Hold on Holds the current graph on the screen Legend Puts a legend on the current screen Margin Computes the gain margin, phase margin , and associated crossover frequencies from frequency response data Bode Generates bode frequency response plots Feedback Computes the feedback interconnection of two systems Trapz Computes the integration value For Generate loop structure End End of loop generated D.2 Appendix D MATLAB Program D.2 Open Loop Programs % Matlab program % for dynamic behavior of open loop with plotting % dynamic behavior of open loop between Tco vs. mh % define hot water flow rate (mh) (kg/sec) at steady state mh=[ value of hot water flow rate at steady state]; % define the transfer function between input mh with outputs Tco with % delay time by using pade function num=[value of nominator]; den=[value of denominator]; [numdt,dendt]=pade (value of delay time, number of approximation); % apply series function [nump,denp]=series (num,den,numdt,dendt); Gp=tf(nump,denp); % where Gp is transfer function between Tco & mh % define outlet cold temperature(C) at steady state (Tco_ss) Tco_ss= value of outlet cold water Temperature(C) at steady state [Tco, x, t]=step (nump,denp); % plotting Tco vs. mh at step response figure (1) plot (t,Tco+Tco_ss ,'k') ylabel ('Outlet Cold Water Temperature (C)') xlabel ('Time(sec)') hold on % or plotting Tco vs. mh at multi-step response [Tco, x, t]=step (value of step1*manuipulated variable*nump,denp); plot (t,Tco+Tco_ss ,'m') [Tco, x, t]=step (value of step2*manuipulated variable*nump,denp); plot (t,Tco+Tco_ss ,'g') D.3 Appendix D MATLAB Program [Tco, x, t]=step (value of step3*manuipulated variable*nump,denp); plot (t,Tco+Tco_ss ,'r') [Tco, x, t]=step (value of step4*manuipulated variable*nump,denp); plot (t,Tco+Tco_ss ,'c') [Tco, x, t]=step (value of step5*manuipulated variable*nump,denp); plot (t,Tco+Tco_ss ,'y') [Tco, x, t]=step (value of step6*manuipulated variable*nump,denp); plot (t,Tco+Tco_ss ,'b') legend(' value of step1',' value of step2',' value of step3',' value of step4', ' value of step5', ' value of step6',2) D.3 Close Loop Programs D.3.1 Ziegler-Nichols Method % Control tuning in the PHE. % by using Ziegler-Nichols method (bode diagram). % define the transfer function of process (PHE) with delay time num=[value of nominator]; den=[value of denominator]; [numdt,dendt]=pade (value of delay time, number of approximation); % apply series function [nump,denp]=series (num,den,numdt,dendt); Gp=tf(nump,denp); % where Gp is transfer function of process with delay time % define the transfer function of measurment numm=[ value of nominator]; denm=[ value of denominator]; Gm=tf(numm,denm) % define the transfer function of control valve numv=[ value of nominator]; D.4 Appendix D MATLAB Program denv=[ value of denominator]; Gv=tf(numv,denv) % apply series function [numvp,denvp]=series(numv,denv,nump,denp) Gvp=tf(numvp,denvp) % where the Gvp=GvGp % specify the frequency range w=logspace(-3, 4,100); [Gm,pm,w]= margin (Gp); % where Gm is the gain margin, pm is the phase margin, w is the frequency. Gmdb=20*log10(Gm) figure(1) bode(Gp,'b-') % % calculation the adjusted parameter of controller (PI) Ku=Gmdb; % where ku is ultimate gain Pu=(2*pi)/w; % where Pu is ultimate period of sustained cycling (sec/cycle) kc=Ku/2.2 ti=Pu/1.2 numc=[kc*ti kc]; denc=[ti 0]; Gc=tf(numc,denc) %or % calculation the adjusted parameter of controller (PID) Ku=Gmdb; %where ku is ultimate gain D.5 Appendix D MATLAB Program Pu=(2*pi)/w; % where Pu is ultimate period of sustained cycling (sec/cycle) kc=Ku/1.7 ti=Pu/2 td=Pu/8 numc=[kc*ti*td kc*ti kc]; denc=[0 ti 0]; Gc=tf(numc,denc) % apply series function [numol,denol]=series(numc,denc,numvp,denvp); GoL=tf(numol,denol) % where the GOL=GpGcGv % apply feedback function [numcl,dencl]=feedback(numol,denol,numm,denm); TFCL=tf(numcl,dencl) % where the TFCL is T.F. of close loop [y,x,t]=step(numcl,dencl); figure(2) % plotting the step respone of close loop plot(t,y,'k-') xlabel('Time (sec)') ylabel(' Outlet Cold Water Temperature (c)') % % find ISE (integral square error) a=y'; % where a is response values % E=set point value- measured value % where E is the error E=1-a; D.6 Appendix D MATLAB Program SE=E.*E; % where SE is square of the error % use trapz function to calculate the area under the curve ISE= trapz(t,SE) figure(3) plot(t,SE,'r-') xlabel('Time (sec)') ylabel('Square of Error') % or % find ITAE (integral time-weighted absolute error) a=y'; % where a is response values % e=set point value- measured value % where e is the error e=1-a ; E=abs(e); TE=t.*E; % where TE is the time* absolute error % use trapz function to calculate the area under the curve ITAE= trapz(t,TE) figure (3) plot (t,TE,'r-') xlabel('Time (sec)') ylabel('Time* absolute error') D.3.2 Cohen-Coon Method % Control tuning in the PHE. % by using Cohen-Coon method (process reaction curve). % define the transfer function of process (PHE) with delay time D.7 Appendix D MATLAB Program num=[value of nominator]; den=[value of denominator]; [numdt,dendt]=pade (value of delay time, number of approximation); % apply series function [nump,denp]=series (num,den,numdt,dendt); Gp=tf(nump,denp); % where Gp is transfer function of process with delay time % define the transfer function of measurment numm=[ value of nominator]; denm=[ value of denominator]; Gm=tf(numm,denm) % define the transfer function of control valve numv=[ value of nominator]; denv=[ value of denominator]; Gv=tf(numv,denv) % apply series function [numvp,denvp]=series(numv,denv,nump,denp) Gvp=tf(numvp,denvp) %where the Gvp=GvGp [Tco,x,t]=step(nump,denp); figure (1) plot(t,Tco,'r') xlabel('Time(sec') ylabel(' Outlet Cold Water Temperature (c)') hold on % finding the values of k,Tau and td from figure (1) k=[ values of k]; Tau=[ values of Tau 1]; td= values of td; D.8 Appendix D MATLAB Program [Tco,x,t]=step(k,Tau); plot(t+td,Tco,'b') hold off % from figure (1) k= values of k; Tau= values of Tau; td= values of td; % calculation the adjusted parameter of controller(PI) kc=(Tau/(k*td))*(0.9+(td/(12*Tau))) ti=td*((30+((3*td)/Tau))/(9+((20*td)/Tau))) numc=[kc*ti kc]; denc=[ti 0]; Gc=tf(numc,denc) % or % calculation the adjusted parameter of controller(PID) kc=(Tau/(k*td))*((4/3)+(td/(4*Tau))) ti=td*((32+((6*td)/Tau))/(13+((8*td)/Tau))) td=td*((4)/(11+((2*td)/Tau))) numc=[kc*ti*td kc*ti kc]; denc=[0 ti 0]; Gc=tf(numc,denc) % apply series function [numol,denol]=series(numc,denc,numvp,denvp); GoL=tf(numol,denol) % where the GOL=GpGcGv % apply feedback function [numcl,dencl]=feedback(numol,denol,numm,denm); TFCL=tf(numcl,dencl) % where the TFCL is T.F. of close loop D.9 Appendix D MATLAB Program [y,x,t]=step(numcl,dencl); figure(2) % plotting the step respone of close loop plot(t,y,'k-') xlabel('Time (sec)') ylabel(' Outlet Cold Water Temperature (c)') % % find ISE (integral square error) a=y'; % where a is response values % E=set point value- measured value % where E is the error E=1-a; SE=E.*E; % where SE is square of the error % use trapz function to calculate the area under the curve ISE= trapz(t,SE) figure(3) plot(t,SE,'r-') xlabel('Time (sec)') ylabel('Square of Error') % or % find ITAE (integral time-weighted absolute error) a=y'; % where a is response values % e=set point value- measured value % where e is the error e=1-a ; E=abs(e); D.10 Appendix D MATLAB Program TE=t.*E; % where TE is the time* absolute error % use trapz function to calculate the area under the curve ITAE= trapz(t,TE) figure (3) plot (t,TE,'r-') xlabel('Time (sec)') ylabel('Time* absolute error') D.11 Appendix E Experimental Data of Dynamic Behavior Table (E.1) ( ∆ T lm ) vs. (T hi - T ho ) at (m h =0.0497) (kg/sec) and (m c =0.0414) (kg/sec). ( ∆ T lm ) oC (T hi - T ho ) oC 11.7124 2.1000 12.4615 2.3000 13.1504 2.4000 13.8594 2.8000 14.0808 3.0000 14.7251 3.2000 15.4305 3.6000 15.6436 3.8000 Table (E.2) ( ∆ T lm ) vs. (T hi - T ho ) at (m h =0.0579) (kg/sec) and R R R R R R R R (m c =0.0414) (kg/sec). R R ( ∆ T lm ) oC (T hi - T ho ) oC 11.5974 3.1000 12.2385 3.1000 13.0315 3.1000 14.2453 4.3000 14.9433 4.5000 15.1998 4.7000 15.3922 4.0000 15.9925 4.4000 16.6902 4.6000 R R P P R E.1 R R R P P Appendix E Experimental Data of Dynamic Behavior Table (E.3) ( ∆ T lm ) vs. (T hi - T ho ) at (m h =0.0662) (kg/sec) and (m c =0.0414) (kg/sec). ( ∆ T lm ) oC (T hi - T ho ) oC 13.2424 3.2000 13.7804 3.3000 14.5815 3.5000 15.2736 3.6000 15.5923 3.8000 16.2926 4.1000 16.8903 4.4000 17.5907 4.7000 Table (E.4) ( ∆ T lm ) vs. (T hi - T ho ) at (m h =0.0745) (kg/sec) and R R R R R R R R (m c =0.0414) (kg/sec). R R ( ∆ T lm ) oC (T hi - T ho ) oC 11.6929 2.8000 12.3348 2.9000 13.0269 3.0000 13.7758 3.2000 14.4721 3.4000 15.2158 3.5000 15.4379 3.8000 16.1351 4.0000 15.8527 3.6000 R R P P R E.2 R R R P P Appendix E Experimental Data of Dynamic Behavior Table (E.5) ( ∆ T lm ) vs. (T hi - T ho ) at (m h =0.0828) (kg/sec) and (m c =0.0414) (kg/sec). ( ∆ T lm ) oC (T hi - T ho ) oC 11.9372 2.5000 12.7244 2.5000 13.5202 2.6000 14.2242 2.9000 14.4548 3.1000 15.1036 3.4000 15.8057 3.7000 16.4544 4.0000 16.0533 3.4000 Table (E.6) ( ∆ T lm ) vs. (T hi - T ho ) at (m h =0.091) (kg/sec) and R R R R R R R R (m c =0.0414) (kg/sec). R R ( ∆ T lm ) oC (T hi - T ho ) oC 11.3707 2.6000 12.2729 2.7000 12.7022 2.9000 12.4935 2.4000 13.2999 2.6000 13.9951 2.8000 14.2865 3.0000 14.9194 3.1000 15.7458 3.6000 R R P P R E.3 R R R P P Appendix E Experimental Data of Dynamic Behavior Table (E.7) ( ∆ T lm ) vs. (T hi - T ho ) at (m h =0.0993) (kg/sec) and (m c =0.0414) (kg/sec). ( ∆ T lm ) oC (T hi - T ho ) oC 12.7090 3.0000 12.8708 2.3000 13.6666 2.4000 14.4713 2.6000 14.6020 2.9000 15.3575 3.2000 16.1065 3.4000 16.8083 3.7000 Table (E.8) ( ∆ T lm ) vs. (T hi - T ho ) at (m h =0.1076) (kg/sec) and R R R R R R R R (m c =0.0414) (kg/sec). R R ( ∆ T lm ) oC (T hi - T ho ) oC 11.5651 2.3000 12.1537 2.5000 12.8094 2.9000 12.9714 2.2000 13.7299 2.5000 14.0129 2.7000 14.6616 3.0000 15.3575 3.2000 R R P P R E.4 R R R P P Appendix E Experimental Data of Dynamic Behavior Table (E.9) ( ∆ T lm ) vs. (T hi - T ho ) at (m h =0.1159) (kg/sec) and (m c =0.0414) (kg/sec). ( ∆ T lm ) oC (T hi - T ho ) oC 12.1383 2.3000 13.0425 2.4000 13.7379 2.6000 14.5414 2.8000 15.2441 3.1000 16.0791 3.0000 16.1652 3.5000 16.3878 2.8000 Table (E.10) The values of overall heat transfer coefficient (U) as a function of hot water flow rate (m h ). R m h (kg/sec) U (w/m2.oC) 0.0497 1077.605 0.0579 1653.626 0.0662 1695.345 0.0745 1809.100 0.0828 1846.360 0.0910 1985.241 0.0993 2064.224 0.1076 2198.045 0.1159 2286.572 R R P E.5 P P P R Appendix E Experimental Data of Dynamic Behavior Table (E.11) System parameters for different step change. Step size % k τ (sec) 20% 103.62 36.45 50% 96.58 35.71 80% 95.82 32.78 100% 94.96 30.53 120% 93.9 30 135% 93.3 29.2 Fig. (E.1) Temperature difference (T hi - T ho ) as a function of ( ∆ T lm ) for (m h =0.0497) (kg/sec) and (m c =0.0414) (kg/sec). E.6 Appendix E Experimental Data of Dynamic Behavior Fig. (E.2) Temperature difference (T hi - T ho ) as a function of ( ∆ T lm ) for (m h =0.0579) (kg/sec) and (m c =0.0414) (kg/sec). Fig. (E.3) Temperature difference (T hi - T ho ) as a function of ( ∆ T lm ) for (m h =0.0662) (kg/sec) and (m c =0.0414) (kg/sec). Fig. (E.4) Temperature difference (T hi - T ho ) as a function of ( ∆ T lm ) for (m h =0.0745) (kg/sec) and (m c =0.0414) (kg/sec). E.7 Appendix E Experimental Data of Dynamic Behavior Fig. (E.5) Temperature difference (T hi - T ho ) as a function of ( ∆ T lm ) for (m h =0.0828) (kg/sec) and (m c =0.0414) (kg/sec). Fig. (E.6) Temperature difference (T hi - T ho ) as a function of ( ∆ T lm ) for (m h =0.091) (kg/sec) and (m c =0.0414) (kg/sec). Fig. (E.7) Temperature difference (T hi - T ho ) as a function of ( ∆ T lm ) for (m h =0.0993) (kg/sec) and (m c =0.0414) (kg/sec). E.8 Appendix E Experimental Data of Dynamic Behavior Fig. (E.8) Temperature difference (T hi - T ho ) as a function of ( ∆ T lm ) for (m h =0.1076) (kg/sec) and (m c =0.0414) (kg/sec). Fig. (E.9) Temperature difference (T hi - T ho ) as a function of ( ∆ T lm ) for (m h =0.1159) (kg/sec) and (m c =0.0414) (kg/sec). E.9 Appendix F Calculation of Overall Heat Transfer Coefficient (U) Calculation of Overall Heat Transfer Coefficient (U) The rate of heat transferred through the exchanger Q is given by: ……….. (F.1) Q = U A ∆Τ lm Where: A: Area of heat transfer (m2). P P o ∆ T lm : Logarithmic mean temperature difference ( C). R R P (∆Τ ) = (Τ − Τ ) − (Τ − Τ ) ln Τ − Τ Τ −Τ ci hi lm hi co ho co ci ho P ……….. (F.2) Then equation (F.1) become: ( − )− ( − ) Q = UA Τ Τ −Τ Τ Τ Τ ln − Τ Τ hi co ho ……….. (F.3) ci hi co ho ci This amount of heat is equal to the enthalpy lost by the heating fluid, which is the same as the enthalpy gained by the process fluid providing that there are negligible heat losses. Therefore, Q = m C (Τ − Τ ) = m C (Τ − Τ ) h hi ph ho c co pc ……….. (F.4) ci Since both fluids are water, it is reasonable to assume that the specific heat C ph . = C pc = C p Thus: m (Τ − Τ ) = m (Τ − Τ ) = hi h ho c co ci hi ……….. (F.5) p (Τ − Τ ) − (Τ − Τ ) − ln Τ − Τ Τ Τ UA ∴ m (Τ − Τ ) = C h Q C hi ho p co ho ci hi co ho ci ……….. (F.6) Or: ……….. (F.7) (Τ − Τ ) = UA ∆Τ mC hi ho lm h p The above equation shows a linear relationship between (T hi - T ho ) R R R R and ∆ T lm with slop UA from which U can be determined. From each R R mC h p figure a single value of the overall heat transfer coefficient (U) was F.1 Appendix F Calculation of Overall Heat Transfer Coefficient (U) determine for each hot water flow rate (m h ) and the complete set values can R be found in table (E.10) in appendix (E). F.2 R Fig. (3.2) Schematic diagram of the experimental rig.