International Journal of Emerging Technology and Advanced Engineering Website: www.ijetae.com (ISSN 2250-2459, Volume 2, Issue 6, June 2012) Fuzzy Modelling of Voltage Standing Wave Ratio using Fuzzy Regression Method T. D. Dongale1, T .G. Kulkarni2, R. R. Mudholkar3 1,2,3 Department of Electronics, Shivaji University, Kolhapur The fuzzy reasoning is broadly classified into two methods: first the Direct method and second the Indirect method. The direct method includes Mamdani’s method, Takagi and Sugeno’s fuzzy modeling and Simplified method. The direct methods are popular due to their simplicity. The indirect methods are based on truth value space. The indirect method has complex reasoning mechanism. In the fuzzy modeling of VSWR we have employed the fuzzy reasoning by Takagi-Sugeno reasoning, where consequence part of the rule is represented by linear function. Abstract— The transmission line or waveguide exhibit the nonlinear characteristics of VSWR (Voltage Standing Wave Ratio) pattern. An attempt is made to model the VSWR pattern characteristics using method of Fuzzy Reasoning and Fuzzy Regression Method. The very idea of fuzzy reasoning by linear function has been explored in modelling the VSWR pattern. The results show a great resemblance between practical results and that obtained using fuzzy based VSWR model. The VSWR pattern of any transmission line or waveguide consists of many other parameters which profoundly affect the transmitting and receiving signal. These include the parameters such as Return Loss, Reflection Coefficient and mismatch loss. The modelling of VSWR pattern is demonstrated by using Sugeno Reasoning Method. This reasoning method consists of output membership function as linear function. The fuzzy regression is used for the reducing the output linear equation. Fuzzy Regression Method reduces the output linear equations from 49 to 7. This also demonstrates the feasibility of using Fuzzy reasoning in the microwave device modelling exhibiting non-linear behaviour. Keywords— Fuzzy Logic, VSWR pattern, Reasoning, Fuzzy Regression, linear function Sugeno I. INTRODUCTION The fuzzy logic in broad sense is the theory that includes fuzzy sets, fuzzy logic, fuzzy reasoning, fuzzy measure, fuzzy relation etc. Its main objective lies in the modeling of complex, nonlinear dependency that exists between input and output variables of a system. Further it aims to represent system operation knowledge in the form of linguistic rules [1]. The basic elements of fuzzy system are shown in figure.1. Amongst them reasoning forms a key element. Reasoning is a process by which you can reach a conclusion after consulting all the facts i.e. reasons behind the decision. This paper demonstrates how fuzzy model of VSWR can be realized based on fuzzy reasoning. Fuzzy reasoning is a process in which given a value of input antecedent variable, yields the values of consequent output variable based on imprecise and non-linear dependency formulated in the rule base [2, 3]. Figure.1. Fuzzy Inference System II. FUZZY REASONING USING LINEAR FUNCTION The Direct method of fuzzy reasoning has following difficulties such as, 1. As premise variable increases, the number of rules increases exponentially. 2. As number of rule increases the time required to construct the rule also increases along with complexity. 3. If number of premises is too large then it becomes difficult to keep track of mapping between input and output premises. To overcome these problems the fuzzy reasoning by linear function is adopted for consequence part. 21 International Journal of Emerging Technology and Advanced Engineering Website: www.ijetae.com (ISSN 2250-2459, Volume 2, Issue 6, June 2012) This greatly reduces the number of fuzzy rules and accelerates the reasoning process. However it needs a skillful off-line modeling work to identify the functional relationship between input and output variables forming the consequence part of fuzzy rule [1, 3]. A. VSWR Model - A Fuzzy Approach The VSWR is dependent of the reflection coefficient, mismatch loss, return loss, the dependency however is nonlinear. Hence for modelling the VSWR a Sugeno method of Fuzzy reasoning has been used. This approach involves the construction of fuzzy sets for antecedent part and linear function for consequence part to model nonlinear mapping of VSWR with reflection coefficient, mismatch loss and the return loss. The fuzzy model of VSWR relies on the linear equation of the form: y = mx + c (1) The whole range of VSWR has been partitioned into seven fuzzy sets labelled as VERY LOW, LOW, MEDIUM, MEDIUM HIGH, HIGH, VERY HIGH and BIG. The ANFIS EDIT tool box of MATLAB creates the required number of input-out membership functions, rules and linear equations. For modelling of VSWR fuzzy rule by triangular membership function for antecedent part and linear function for consequence part has been employed. The rules are defined here on universe of discourses with respect to reflection coefficient, mismatch loss and the return loss. The observations of VSWR used for modelling are shown in figure 2, 3 and 4. Figure.3.Relationship between Reflection Coefficients (Γ) and VSWR pattern Figure.4.Relationship between Mismatch loss and VSWR pattern VSWR is the ratio of the maximum voltage to the minimum voltage in the standing wave on a transmission line [7]. Standing waves are the result of reflected RF energy. When the reflections on the line approaches zero the maximum power may be transmitted. Reflections occur at any place where the impedance of the transmission line changes. Usually this phenomenon happens on a radio or radar transmission line. If all the energy gets reflected (for example, by an open or short circuit) at the end of the line, then none gets absorbed producing a perfect ‘standing wave’ on the line. This is undesirable situation. If the reflected wave is not as strong as the forward wave, then some ‘standing wave’ pattern will be observed, but the nulls will not be as deep or the peaks as high as for a perfect reflection (or complete mismatch). Therefore, any ‘standing wave’ is an indication of an imperfect condition, with part of the power meant for radiation being returned because of a mismatch. Figure.2.Relationship between Return loss and VSWR pattern 22 International Journal of Emerging Technology and Advanced Engineering Website: www.ijetae.com (ISSN 2250-2459, Volume 2, Issue 6, June 2012) If there is no mismatch with only a forward travelling wave, then there is no standing wave; i.e., the voltage at any point on the line is the same as it is everywhere else. The intensity of the standing wave is referred to as the ‘Voltage Standing Wave Ratio’ abbreviated as VSWR. [5] There are four quantities that describe the effectiveness of transferring power from a line to a load or antenna: the VSWR, the reflection coefficient, the mismatch loss and the return loss [6]. These are all inter-related with each other. The voltage component of a standing wave in a uniform transmission line consists of the forward wave (with amplitude, Vf) superimposed on the reflected wave (with amplitude, Vr). Reflections occur as a result of discontinuities, such as an imperfection in an otherwise uniform transmission line, or when a transmission line is terminated with other than its characteristic impedance. If the feed line has no loss, and matches both the transmitter output impedance and the generator input impedance, then the maximum power will be delivered to the load. In this case the VSWR will be 1:1 and the voltage and the current will be constant over the whole length of the feed line. Return loss is a measure in dB of the ratio of power in the incident wave to that in the reflected wave, and we define it to have a negative value. The higher the return loss, minimum is the less power loss. Return loss = 10 log10 [Pr/Pi] =20 log10 [Er/Ei] The output variable is VSWR. Figure.5 depicts snap shot of FIS-Editor window for design structure of VSWR model. Figure.5. FIS editor for VSWR model (2) Figure.6.Input membership function (Return loss) editor for VSWR model Also of considerable interest is the mismatch loss. This is a measure of how much the transmitted power is attenuated due to reflection. It is given by the following relation, Mismatch Loss = 10 log (1 -ρ 2) (3) The reflection coefficient Γ is equal to, Γ= (ZL-Z0)/ (ZL l+Z0) (4) The load impedances ZL and characteristic impedance Zo of the transmission line are in ohm. The reflection coefficient magnitude, |Γ| or ρ, is the ratio of the amplitude of the reflected wave to the amplitude of the incident wave at the junction of a transmission line and the terminating impedance. |Γ| has a value lying between 0 and 1. The |Γ| = 0 means the line is perfectly matched, and a value of 1 means that the line is either shorted or open-circuit. [6-7] B. Fuzzy Inference System (FIS) for VSWR modelling The fuzzy inference systems associated with input and output variables create a fuzzy mapping between the design variables. For simple consequence part Sugeno method is good choice. The input variables are return loss, reflection coefficient and mismatch loss. For a better result of modelling, we have chosen five triangular and two trapezoidal membership functions. The knowledge base pertaining to the VSWR modelling is formulated in terms of fuzzy inference rules, which are supported by the data-base. The knowledge-base thus comprises a Data-base and a Rule-base. The data-base provides the required information to fuzzification, rule-base and defuzzification modules. The information includes, the membership functions representing meanings of linguistic values of input and output variables in addition to Labels, shapes, slopes and domain-ranges. The Figure.6.depicts snap shot of Membership function-Editor window for input variable ‘Return Loss’. The ranges of membership function considered for Return loss are as followsμVS (x) = L (x,0,1.8,2.2) μSM (x) = ^ (x,3.6,7,10.4) μM (x) = ^ (x,7,10.4,14) μL (x) = ^ (x,10.4,14,17.8) μVL (x) = ^ (x,14,17.8,21) μB (x) = Γ (x,18,20,20) 23 International Journal of Emerging Technology and Advanced Engineering Website: www.ijetae.com (ISSN 2250-2459, Volume 2, Issue 6, June 2012) The symbol ‘L’ stands for left bounded membership function, the ‘Γ’ for right bounded membership function and the ‘^’ symbol indicate that membership function centred triangular membership function. μSM (x) = ^ (x,1.3,2.5,3.8) μM (x) = ^ (x,2.5,3.8,5) μL (x) = ^ (x,3.8,5,6.2) μVL (x) = ^ (x,5,6.2,7.5) μB (x) = Γ (x,6.4,7,7.5) This modelling is based on TSK method in which consequence part in the form of linear equation of the form: Y= mx+C. For modelling Adaptive Neuro-Fuzzy Inference System (ANFIS) tool from MATLAB is being used. The Neuro-adaptive learning method works similarly to that of neural networks. Neuro-adaptive learning techniques provide a method for the fuzzy modelling procedure to learn based on the information about a data set. Fuzzy Logic Toolbox software computes the membership function parameters that best allow the associated fuzzy inference system to track the given input/output data. The Fuzzy Logic Toolbox function that accomplishes this membership function parameter adjustment is called ANFIS. The ANFIS function can be accessed either from the command line or through the ANFIS Editor GUI as the functionality of the command line function ANFIS and the ANFIS Editor GUI is similar in nature. In this tool the decision making such as rule formulation, number of rule etc. is done automatically according to data provided in the form of .MAT format. Even though it can create output linear equation, but we can edit this equation according to need based tuning [8]. For modelling of data on VSWR, first of all the data set is converted into linear equation of the form (1). If ‘y’ is a function of more than one independent variable, the matrix equations that express the relationships among the variables are expanded to accommodate the additional data. This is called multiple regressions. A model of this data is formed, where multiple regressions solve the modelling equations for unknown coefficients β1, β2 and β3 by minimizing the sum of the squares of the deviations of the data from the model (least-squares-fit technique). By Constructing and solving the set of simultaneous equations by forming the Vander monde matrix [X], and solving for the parameters Figure.7.Input membership function (Reflection Coefficient) editor for VSWR model The Figure.7.depicts snap shot of Membership functionEditor window for input variable ‘Reflection Coefficient’. The ranges of membership function considered for Reflection Coefficient are as followsμVS (x) = L (x,0,0.1,0.14) μS (x) = ^ (x,0,0.15,0.30) μSM (x) = ^ (x,0.15,0.30,0.45) μM (x) = ^ (x,0.30,0.45,0.60) μL (x) = ^ (x,0.45,0.60,0.75) μVL (x) = ^ (x,0.60,0.75,0.90) μB (x) = Γ (x,0.78,0.85,0.90) computed as flows-[8] X = [ones (size(x1)) x1 x2 x3] The coefficient vector is: Figure.8.Input membership function (mismatch loss) editor for VSWR model The Figure.8.depicts snap shot of Membership functionEditor window for input variable ‘Mismatch Loss’. The ranges of membership function considered for Mismatch loss are as followsμVS (x) = L (x,0,0.8,1.2) μS (x) = ^ (x,0,1.3,2.5) a = X/y β1 = 0.0548 β2 = -7.9246 β3 =3.2493 Here x1, x2, x3 are the data set of Return loss, Reflection coefficient and Mismatch loss respectively. 24 International Journal of Emerging Technology and Advanced Engineering Website: www.ijetae.com (ISSN 2250-2459, Volume 2, Issue 6, June 2012) From the data set we get a one linear equation which is given by least-squares method. The least-squares fit model of the data is as given by equation (5), Y= β0 + β1 x1 + β2 x2+ β3 x3 Y= 1.5396 + 0.0548 x1 – 7.9246 x2 + 3.2493 x3 (5) Where, β0 is called as intercept and β1, β2, and β3are called as coefficient of linear equation. To validate the model, we find the maximum of the absolute value of the deviation of the data from the actual model. Y = X*a; MaxErr = max (abs (Y - y)) MaxErr = 0.13 Where, y is called as observed value and Y called as estimated value. The error is 0.13 between observed and estimated value which is very small. It means that the equation (5) models the system’s behaviour. The intercept part and coefficient of equation gives the actual behaviour of small set of data. Hence instead of using different output linear equations for individual data in TSK method [10], we can use only one regression equation for set of data. By using the Fuzzy Regression Method, we can eliminate unnecessary linear equations. But only one equation may give some sort of error. For minimizing the error, we can create suitable linear equation by fuzzy regression method through number of iterations based on trial and fit process. The seven different linear equations (6 to 12) give zero error for data set. Hence it is suitable for modelling of VSWR pattern. By fuzzy regression method only seven linear equations instead of forty nine equations are sufficient to model the VSWR. The linear equations and rule base are as follows which satisfies the actual behaviour of observed data. If VSWR is VERY LOW Then Y= 0.9996 - 0.020 x1 + 2.5791 x2 + 0.0865 x3 (6) If VSWR is LOW Then Y= -1.3292 + 0.1473 x1+4.1237 x2 + 1.1046 x3 (7) If VSWR is MEDIUM Then Y= 9.9692 - 0.7692 x1 – 9.2308 x2 + 1.5385 x3 (8) If VSWR is MEDIUM HIGH Then Y=-3.5205 + 0.6849 x1– 1.3699 x2+2.7397 x3 (9) If VSWR is HIGH Then Y= -31.00 + 0.00 x1 +50.00 x2 + 0.00 x3 (10) If VSWR is VERY HIGH Then Y= -23.7059 + 5.8824 x1 – 5.8824 x2 + 5.8824 x3 (11) If VSWR is BIG Then Y= -29.97 + 5.8455x1 – 1.7745x2 + 6.1587x3 (12) Figure.9.Output membership function editor for VSWR model Figure.10.Rule viewer for VSWR model III. CONCLUSIONS The process of data reconstructed pertaining to VSWR characteristics using Sugeno Fuzzy Inference is shown in figure 10. Comparison with actual data and observed rule viewer graph shows resemblance between practical VSWR characteristics and characteristics generated by fuzzy model of VSWR. Idea of fuzzy modeling of VSWR is though simple but demonstrates powerful application of fuzzy logic in device modeling based on imprecise data. The same idea can be extended to highly non-linear and complex electronic devices. 25 International Journal of Emerging Technology and Advanced Engineering Website: www.ijetae.com (ISSN 2250-2459, Volume 2, Issue 6, June 2012) The Fuzzy Regression is very useful for creating linear equation when set of data is very large. Instead of using different output linear equations for individual data in TSK method, we can use only one regression equation for set of data. By using this Fuzzy Regression Method we can eliminate unnecessary linear equations. These sets of linear equation for VSWR pattern are discussed from equation number 6 to equation number 12. Also fuzzy regression gives unique linear equation for whole system in some tolerable error of 0.13. This validates the successful implementation of Fuzzy Reasoning by TSK method and Fuzzy Regression Method in modeling the VSWR. REFERENCES [1] Kazuo Tanaka, (1996), an Introduction to Fuzzy Logic for Practical Applications, Department of Mechanical Systems Engineering, Kanazawa University, Japan, Springer Publication page no. 81 - 119 [2] Stamations V. Kartalos, (2005), Understanding neural network and Fuzzy logic, basic concept and Application, AT & T Bell lab, IEEE Neural Network council, sponsor, Prentice Hall of India [3] R.R. Yagar & D.P. Filev, (1994), Essential of fuzzy modelling and Control, Institute of lona college, A Wiley – Interscience Publication [4] Bart Kosko, (2001), Neural network and fuzzy system- a dynamic Approach to machine intelligence, University of south California, Prentice hall of India, page no. 263-296 [5] Zouhair Benmoussa and Don Barrick, (April 2006), the Effects of VSWR on Transmitted Power [6] R.E. Collien, (1987), Fundamental of Microwave Engineering, page No. 82-92 [7] S.Y. Liao, (2006), Microwave Device and Circuit, page no. 79-100 [8] http://www.mathworks.com [9] http://www.microwave101.com/microwave.calculator/VSWR Calculator [10] Tukaram .D. Dongale, T .G. Kulkarni, P. A. Kadam, R. R. Mudholkar, (2011), Fuzzy Model of Thermistor, International Journal of Applied Engineering Research, Dindigul, ISSN 0976-4259 Volume 2, No.1 26