Caledonian Journal of Engineering Volume 04, Number 01 January - June 2008 Published by Caledonian College of Engineering, Post Box 2322, CPO Seeb, Postal Code 111, Sultanate of Oman. General Information Caledonian Journal of Engineering is published bi-annually by Caledonian College of Engineering. The journal endeavors to publish high quality, peer reviewed research papers in the field of Science and Technology to encourage research and scholarly activities in Oman, and also in the Gulf. All correspondence and contributions may be sent to: Associate Editor, Caledonian Journal of Engineering Caledonian College of Engineering Post Box 2322 CPO Seeb 111 Sultanate of Oman Email: kaleel@caledonian.edu.om. Information for Authors The topics covered by the journal include: • Research / Technical papers in the filed of Mathematics, Civil Engineering, Mechanical Engineering, Chemical Engineering, Process Operation and Maintenance Engineering, Electrical Engineering, Electronic Engineering, Communication Engineering and Computer Engineering Responsibilities of Authors The authors are responsible for the originality of their papers and their scientific correctness. Guidelines for paper submission All papers must be the author’s original work. It is essential that each manuscript should be accompanied by an abstract of maximum of 200 words. All authors’ contact addresses, emails, telephone numbers and fax numbers should be included in the paper. The title of a paper must be restricted to 50 words. Please refer to the template for other information regarding the submission of papers. (The template is available in the inner back cover page). The Editorial Board has the right to accept or reject a paper, and to make necessary modifications. The board’s decision shall be held final. Caledonian Journal of Engineering Editorial Board Editor-in-Chief Dr.Syed Anisuddin Editors Dr.Feroz Shaik Dr.Syed Mohammed Rizwan Associate Editor Dr.K.P.Ramachandran Caledonian College of Engineering, Oman Dr.K.V.Gangadharan National Institute of Technology, Surathkal, India Dr.Gulshan Taneja M D University, Rohtak, Haryana, India Prof.James Sommerville Glasgow Caledonian University, Scotland, United Kingdom H.Kalilur Rahman Prof.Mehmet A. Hastaoglu Department of Energy Systems, GYTE, Kocaeli, Turkey Members Dr.Faris Salman K.P. Mansoor Ali Prof.Gautam Datt Dr.G.Prabhakaran Advisory Committee Dr.A.Arunagiri Multimedia University, Malaysia Dr.Brian Stewart Glasgow Caledonian University, Scotland, United Kingdom Prof.Mohsen Morad Sherif College of Engineering, Al Ain, UAE Dr.S.R.R.Senthil Kumar Higher College of Technology, Oman Prof.VSRK.Prasad Andhra University, Visakhapatanam, India Dr.Xavier Fernando Ryerson University, Canada From the Editor-in-Chief’s desk……….. It is with great pleasure that we release Volume 04, Number 01 edition of “Caledonian Journal of Engineering”. Six research papers encompassing various fields of engineering and science are included, and it is particularly gratifying to note that one paper is co authored by faculty of GCU, UK. Papers from AITM Aurangabad and PSG Coimbatore, India are highly appreciated. It is also worth mentioning that the future issues of this journal will bear the International Standard Serial Number as ISSN1999:9496. My thanks go to all the contributors and the editorial team for their consistent hard work. We also welcome contributors from leading higher educational institutions, so that the ideas can be freely exchanged and knowledge is disseminated to all. Dr Syed Anisuddin Editor – in - Chief Caledonian Journal of Engineering Volume04, Number 12, January- June 2008 PAPERS Caledonian Journal of Engineering Volume04, Number 01, January - June 2008 VIRTUAL CELLS FOR MANUFACTURING SYSTEMS UNDER TURBULENT ENVIRONMENT – A REVIEW OF THRUST AREAS R.V.Murali Department of Mechanical & Industrial Engineering, Caledonian College of Engineering, Sultanate of Oman ABSTRACT Cellular manufacturing systems (CMS), since its conceptual introduction about two decades ago, have proved to be more efficient than traditional job shop functional layout systems for a certain range of parameters and datasets. The concept of CMS is based on the philosophy of Group Technology (GT) where the parts, to be machined, are grouped on the basis of similarity in terms of resources requirements. In CMS, existing resources are separated into machine groups or cells that would encircle machineries to process a family of parts. Thus, each cell is responsible for production of a particular family of parts and every time new cells are physically reformed as and when demand for new products or new planning period arrives or product mix is introduced. This poses a difficulty in physical reorganization of the facilities and it will be even worse when the frequency of reformation of cells is high i.e. turbulent environment. In the recent past, a concept called Virtual Cellular Manufacturing Systems (VCMS) which does not demand physical relocation of machines every time to reform a cell, it is rather a conceptual & logical (virtual) grouping of machines from within current departments in order to produce a family of parts. This paper brings out various thrust research areas and opportunities in terms of design, operations and practical applicability (empirical nature) of virtual cell based manufacturing systems in order to meet turbulent environment. to be rearranged and a new set of cells is to be reformed in order to produce new product variety 1.0 INTRODUCTION or new product mix. So much so, classical cellular manufacturing may not be appropriate if While designing of a manufacturing system for a frequent introduction of new variety of products, particular type of industry, it is essential to look or high product mix variability or frequent into various aspects such as product volume change of operational sequence or dynamic variability, product mix variability, physical routing schedule. shape and size of various parts to be produced, processing times, types of machinery required and availability of human workforce, methods of arranging machineries & materials handling systems (layouts) and a strong market competition. In a typical job shop layout (figure 1), machineries are laid out on the basis of their functionality and the raw materials are flowing from one machine to another in order to get transformed into useable products with complete routing flexibility. On the other hand, in a traditional CMS, machine groups (cells) are physically formed and dedicated to a particular family of parts (figure 2). Each cell is engaged in producing a particular family of parts (Greene TJ, Sadowski PR, 1984 & Molleman E, Slomp J, Rolefes S, 2002). In many occasions, more than one cell is required for a job to completely change into a product even though inter-cell movements are kept to a minimal. When a new product arrives or a different variety of product is to be introduced, currently formulated cells are Di – Departments on the basis of type of machine tool Xi, Yi, Zi – Workforce attached with respective department Figure. 1: Job Shop Layout Very recently, a concept called Virtual Cellular Manufacturing Systems (VCMS) is introduced in which there is no need of physical reorganization of machineries every time to form a cell, it is 1 Caledonian Journal of Engineering Volume04, Number 01, January - June 2008 team work, enhanced responsibility & accountability and increased operator expertise . rather a conceptual (virtual & logical) grouping of machines within the machining area in order to produce a family of parts (figure 3). The major research areas where much of the research so far has been focused on are the design part of VCMS and comparison of performance of VCMS with that of traditional cellular manufacturing systems & functional layouts. Design phase of a VCMS clearly refers to formation of machine cells & part families and eventual dedication of particular cells to a particular family of parts so that all/few objectives such as total materials handling cost/time, reduced exceptional elements, effective manpower utilization and increased productive processing time and reduced number of set ups, setup time and tooling are met with. Typical benefits resulting from VCMS include reduced setup and lead times, reduced work in process inventories, reduced materials handling times, proper organization of tooling, accessories VC1 (F1) Figure. 2: Cellular Layout LOOK VC3 (F3) D1 and jigs and fixtures, improved worker responsibility and satisfaction, improvised team work, enhanced productivity and quality and so on. 2.0 VCMS- A LITERATURE VC2 (F3) D2 D3 INTO VIRTUAL CELLS (FOR PLANNING PERIOD ’T1’) Virtual cell concept was first proposed in 1980s when McLean defined it as an imaginary group of machineries which is no longer identifiable explicitly as cells. This virtual concept of cells is present in the minds of people and controlling elements and systems in a shop floor. VC1 (F1) VC2 (F3) VC3 (F3) D1 D2 Although many researchers have defined VCMSs based on their understanding and experience gained over the years, the simplest definition could be forming a pool of machineries virtually and dedicating them to produce a particular family of products / parts so that total set up time, material handling time and part make-span time are minimized to a greater extent. Frequent reformation of cells owing to (i) dynamic topography of products and product mix, (ii) strong competition in the markets, (iii) inability of traditional cellular layouts and functional layouts (Vakharia AJ, Moily JP, Huang Y, 1999) to meet the varied requirements have resulted in virtual cells concept (Rheault M, Drolet JR, Abdulnour G,1995 & 1996). Other perceived benefits of VCMS include improved human relations & D3 VIRTUAL CELLS (FOR PLANNING PERIOD ’T2’) VCx- Virtual Cell Number Figure. 3: Virtual Cells As mentioned earlier, current research works on Virtual Cellular Manufacturing Systems have been towards design and operational sectors. However, there is a little work done in empirical research and hence large potential is identified in this sector. Inputs from empirical research can significantly improve the realism of search 2 Caledonian Journal of Engineering Volume04, Number 01, January - June 2008 introduction of virtual cellular manufacturing systems in 1978 is explained in the graph (Figure 4) and the nature of work done in Table 1. settings, particularly in the area of design and operations. Further, combining multiple types of research i.e. design and/or operation and/or empirical investigative works could undoubtedly enhance the power of virtual cell concepts in terms of quality and applicability (Kanan VR, Ghosh S, 1996). This graph (figure 4) shows various research themes and solution methodology adopted in each theme. It is evident from the graph that the research thrust for VCMS concept appears to be on the rise and still a lot of scope is identified to further explore in the field. 3.0 OPTIMIZATION PARAMETERS IN VCMS 4.0 SIMULATION PHASE FOR THE RESEARCH OUTCOMES In all VCMS research works, key parameters that are to be optimized are identified along with appropriate resource constraints. Accordingly one or more objective functions are formulated. The outcomes of the optimization could well be pertaining to total materials handling costs, productive machining hours to minimize setup and handling time/costs, inter-cell and intra cell movements of parts, cell independence, labor cross training and cell load balancing and so on. A quick review of such parameters would be really helpful here in order to find the impact directions of research yet to be carried out. Although, adequate magnitude of research was done and is being underway on design and operational sectors of VCMS, empirical research on VCMS has not been given much attention. Therefore, a lot of potential is perceived in pressing the outcomes of the research on design and operation into real time implementation. Also, more analytical and simulation phases for the above areas still remain unexplored. 5.0 SUMMARY From the graph (figure 4) and eventual explanation in Table 1, there are some interesting observations that could be noticed. The current research focused on either design aspects or operational aspects or empirical point of view. There is a scope perceived when these three areas are integrated in terms of parameter ranges, data sets and resource types. Ironically, this would enhance the quality and impact of the research to realize the power of VCMS. In papers (Gert Nomden, Jannes Slomp Nallan C. Suresh, 2006), (Irani SA, Cavalier TM, Cohen PH, 1993), (Kannan VR, Ghosh S, 1996) (Vakharia AJ, Moily JP, Huang Y., 1999), (Suresh NC, Meredith JR., 1994), (Slomp J, 1998), various objectives that were formulated for optimization include maximization of productive output [in terms of machining hours], minimization of total number of additional machines of type m needed for creating independent virtual cells, minimization of number of additional cells to which each worker is assigned, maximization of machine coverage of machines and finally maximization of multifunctionality of workers. 6.0 CONCLUSIONS The extensive efforts, investigation and hard work put in by the researchers in the area of VCMS have given us an ample of opportunities to look into to further our research work. The mathematical models developed so far have undoubtedly led us towards a generalized range of parameters in which VMCs may be effectively and efficiently utilized. However, there is a major need to understand the real industrial scenario and therefore we focus our research efforts into more of empirical nature. In paper (R.V.Murali, D.Ragavesh & G. Prabhakaran, 2007), the objective of minimizing the total materials transfer costs (both inter and intra cellular movements of parts) is considered leading to reduced total manufacturing costs. Solution methodologies employed in various research works in order to solve these objective functions include Goal and Integer Programming, LINGO formulation and Tabu and evolutionary search algorithms. The concise version of the detailed research work done in each year right from the conceptual 3 Caledonian Journal of Engineering Volume04, Number 01, January - June 2008 REFERENCES [1] Gert Nomden, Jannes Slomp and Nallan Suresh (2006) Virtual manufacturing cells: taxonomy of past research and identification future research issues, International Journal Flexible Manufacturing Systems 17, p.71-92. [11] Suresh NC and Meredith JR (1994) Coping with the loss of pooling synergy in cellular manufacturing systems, Management Science, 40(4), p.466–483. C. A of of [12] Vakharia AJ and Moily JP, Huang Y (1999) Evaluating virtual cells and multistage flow shops: an analytical approach, International Journal of Flexible Manufacturing Systems,11, p.291–314. [2] Greene TJ and Sadowski PR.(1984) A review of cellular manufacturing assumptions, advantages and design techniques, Journal of Operation Management 4, p.65–97. [3] Irani SA, Cavalier TM and Cohen PH (1993) Virtual manufacturing cells: exploiting layout design and intercell flows for the machine sharing problem, International Journal of Production Research 31(4), p.791–810. [4] Kannan VR and Ghosh S (1996) A virtual cellular manufacturing approach to batch production, Decision Science 27(3), p.519–539. [5] Kanan VR and Ghosh S (1996) Cellular manufacturing using virtual cells, International Journal of Operation, Prodcution Management 16 (5), p.99-112. [6] Molleman E, Slomp J and Rolefes S. (2002) The evolution of a cellular manufacturing system—a longitudinal case study, Int J Prod Econ, 75, p.305–322. [7] Rheault M, Drolet JR and Abdulnour G. (1995) Physically reconfigurable virtual cells: a dynamic model for a highly dynamic environment, Comput Ind Eng, 29(1-4), p.221– 225. [8] Rheault M, Drolet JR and Abdulnour G (1996) Dynamic cellular manufacturing systems (DCMS), Comput Ind Eng, 31, p.143–146. [9] R.V.Murali, D.Ragavesh and Dr.G. Prabhakaran (2007) A Study, design and optimization virtual cellular manufacturing systems, proceedings of Computer Aided Production Engineering (CAPE) international conference, Glasgow Caledonian University, Glasgow, UK. [10] Slomp J (1998) Design of manufacturing cells: PFA applications in Dutch industry. In: Suresh NC, Kay JM, editors. Group technology and cellular manufacturing. Boston, Dordrecht, London: Kluwer Academic Publishers. p.153– 168. 4 Caledonian Journal of Engineering NO YEAR PUBLISHED 1 1978 2 1982 3 1989 4 1991 5 1992 6 1993 7 1994 8 9 1995 10 1996 11 12 1998 13 14 15 1999 16 17 2000 18 Volume04, Number 01, January - June 2008 Table 1: Nature of work done during evolution of VCMS NATURE OF WORK / REMARKS An informal existence of Virtual Manufacturing Cells reported by Altom. In this effort, an attempt is made to analyze the costs and savings associated wile implementing Group Technology (GT). Formal definition of VCMS was given as a cell no longer identified as a fixed physical group of machines and rather dynamically changing. A survey was taken from user and non-users of cellular manufacturing system. It revealed that most of the firms used a hybrid layout (dedicate equipment and manufacturing cells) due to high relocation cost and high product-demand variability. An analytical model was developed for partitioning work centers in order to create cells (as opposed to traditional job-shop and functional layouts.) The analytical model developed in 1991 was carefully looked into in terms of advantages of it over functional layouts. This paper superseded the long time controversy such as doubts on performance and limited adoption of cellular manufacturing in industries. An attempt is made to redefine the term VMCS exploiting the layout design and intercellular flows when machine sharing is imminent. This has resulted in using a pool of resources rather than dedicating a cell to a particular type of part family. In this work, several similar machines are clustered into process department and several dissimilar machines constitute flow lines. Part scheduling problem is formulated and solved by linear programming with an objective of minimizing total travel distance and lateness of the jobs. This effort addressed various issues such as a loss of pooling synergy associated with introduction of cellular manufacturing. Alternatively, this effort recommended the use of part family oriented scheduling (FLP) in the conventional job shop layouts. This has paved the way later to focus more attention on Virtual Cellular Manufacturing Systems (VCMS). This year has seen a detailed survey of US manufacturing practices being adopted among all make-to-order industries. This has ferreted out an indicator of limited applicability of CMS in US industry. FLP was analyzed and renamed as virtual cells using the family based scheduling to realize the scheduling and setup efficiencies while retaining the job-shop layouts. An attempt to propose cellular manufacturing using virtual cells (FLP oriented job-shop layouts) was made. Since the limited applicability of CMS was envisaged from the US survey, a concept of an adaptable cellular manufacturing system was perceived in principle. Firstly in the VCMS history, dedicating workers to part families and cells is initiated by Khuling. Study of VCMS in a multistage flow shops environment was analyzed using queuing theory by Vakharia. This effort brings out the significance of human beings in VCMS environment and insists the importance of information structure that enables the operation between the people and Virtual Manufacturing Cell A framework was proposed here for classification of manufacturing cells where in virtual cells consists of people and equipment that are dedicated to a part/product family and workflow is dependent on time and information flow. A prototype design of VCMS applicable to Small-to-Medium Scale industries was proposed in this year which has increased the attention formally to the possible use of VCMS in industries. An effort was made to explain how cell design is done in real practice and investigated the need and objectives of implementing manufacturing cells and 5 Caledonian Journal of Engineering 19 2001 20 2002 21 22 2003 23 2004 25 2005 26 2006 methods and organization employed to do so. It elaborated the various constrains of implementations. Part scheduling problem is formulated and solved by Lagrangian relaxation approach with an objective of minimizing tardiness A Genetic Algorithm was developed to solve part scheduling problem with an objective of minimizing total travel distances. An integrated framework for production planning and cell formation was carried out through a 3 step approach. A mathematical model was developed to minimize intercellular flows, tardiness, utilization and throughput and solved using tabu search method. Enhanced version of PFA was analyzed and presented. It resulted in formation of virtual cells which can be both either process oriented or product oriented. Algorithms were written to create VMCs considering the routing data and frequently used machines. An attempt is made to present a frame work for the design of VMCs including labor skills and team accountability. This is intended for maximizing the capacity as efficiently as possible and making VMCs as independently as possible using goal programming approach. This work is an extended research to previously done (2004) to include the labor dimension in addition to machines and workers (called as Dual Resource Constrained (DRC) systems) A complete run-through on Virtual manufacturing cells to bring about various research undertaken and identification of future research directions in order to completely realize the power of VCMS Nature of work index 24 Volume04, Number 01, January - June 2008 Published Year Figure.4: VCMS literature published vs Year 6 Caledonian Journal of Engineering Volume04, Number 01, January - June 2008 REMOVAL OF COPPER IONS FROM EFFLUENTS USING COCONUT SHELL COKE IN A FIXED BED ADSORBER S Feroz and Shah Jahan Department of Mechanical and Industrial Engineering, Caledonian College of Engineering, Sultanate of Oman ABSTRACT The removal of copper metal ions from effluents using charcoal coke in fixed bed adsorption column is investigated. The effects of various parameters like flow rate, particle size and initial concentration are studied. It was observed that an increase in flow rate, initial concentration decreases the time required for saturation of bed whereas with an increase in particle size of the adsorbent the time of saturation of the bed increased. KEYWORDS: Coconut Shell Coke, Fixed Bed Adsorber, Saturation Curve, Adsorbent. coagulation, reverse osmosis, electro dialysis, ion exchange, and adsorption have been employed for removal of metal ions from effluents. 1.0 INTRODUCTION Water, a universal solvent is in true sense “the elixir of life”. With diminishing sources of clean water, its management has become a vital issue. The constituents in natural water exhibit properties that may be classified as either conservative or non- conservative. The former relates to concentration of species that remains at relatively constant ratio to each other throughout the system. This applies mostly to more abundant lighter elements such as sodium, potassium and calcium. The non-conservative behavior is shown by other metals such as Fe, Cu, Zn, Pb, Hg which varies in concentration depending on variables such as position, time, temperature and most importantly, biological activity. The present work is aimed at removing the copper ions from effluent water using the technique of adsorption. Adsorption involves contact of solids with either liquids or gases and the mass transfers in the direction of fluid to solid. The adsorption operation involves the property of certain solids, which selectively adsorb specific substances from solution on to their surfaces. All adsorption processes are exothermic and adsorption on solid surfaces may be classified on the basis of magnitude of the energy of adsorption. Adsorption depends on temperature, pressure, surface structure, size and porosity. With rise in pressure or decrease in temperature, the adsorption capacity increases. As the particle size decreases, the interfacial area increases thereby enhancing adsorption but it is often restricted due to clogging, if the adsorbent particle size is the lowest. Modern industries require large quantities of fresh water. A substantial quantity of this water is discharged as waste water or effluent, which contain large amount of pollutants. Pollutants from industries involves proteins, fats, carbohydrates, inorganic chemicals, salts of metals like Ag, Cd, Cu, Fe, Ni, Zn etc. Copper is found in effluents from some major industries like chlor alkali, electroplating, paints and dyes, petroleum refining, fertilizers etc. The adsorption operations can be batch, semibatch or continuous. Batch operations are generally conducted when small amounts are to be treated and the equilibrium distribution depends on the time of contact. The semi-batch operations are generally conducted in fixed bed towers and in such towers the solid is stagnant while the fluids are continuous. In continuous counter current operations both solid and fluid are continuously brought into contact with each other in the opposite direction with a continuous flow of two streams. This operation is used when Pollutants cause direct toxicity both to human and other living beings due to their presence beyond specific limits. The trust of environmental research is to adopt suitable techniques either to prevent the metal pollution or to reduce it to very low levels. Various techniques such as chemical precipitation, 7 Caledonian Journal of Engineering Volume04, Number 01, January - June 2008 obtained on all the metal ions except copper followed Fruendlich isotherms. the cost of adsorbent is cheap (McCabe W.L., Smith J.C., Peter Harriot). Bauxite, molecular sieves, alumina, silica gel, peat, coconut shell coke, bagasse, flyash, etc., are commonly used adsorbents (Treybal R.E.). This investigation is confined to fixed bed adsorber, where the copper metal ions present in the effluents are adsorbed on to coconut shell coke adsorbent and the effect of flow rate, initial concentration and particle size are studied. Seco et al., (Seco Aurrora, Marzal Paula, Gabaldon Carmen and Ferror Jose, 1999) conducted studies for adsorption of cadmium and zinc on activated carbon, influence of pH, cation concentration and adsorbent concentration. The single adsorption of cadmium and zinc from aqueous solutions were investigated on granular activated carbon in a wide range of experimental conditions, pH, metal and carbon concentration. Activated carbon is efficient as sorbent for both metals. Metal removal increased with pH and carbon concentration, and decreased with the initial metal concentration. The adsorption processes were modeled using the surface complex formation triple layer model (TLM) and its parameters were determined. Modelling was performed assuming a single surface stiochiometry, which successfully predicted cadmium and zinc removal in all the experimental conditions. 2.0 LITERATURE REVIEW Gupta et al., (Gupta Vinod K, Ali and Imran, 2000) studied the utilization of bagasse, fly ash for the removal of zinc from waste water. Bagasse, fly ash, a waste product in sugar industries, has been converted into an inexpensive and effective adsorbent. The product was characterized by different chemical and physical methods and has been used for the removal of zinc from waste waters. Various parameters such as pH, adsorbent dosage, initial metal ion concentration, temperature, particle size etc., were optimized. Zinc was adsorbed by the adsorbent up to 90 to 95% in batch and column experiments. 3.0 EXPERIMENTAL SET-UP The experimental set-up shown in the schematic diagram (Figure.1) consists of a storage tank of mild steel with dimensions 24x10x12 inches. A submerged pump of capacity 0-400 lph is used to pump the feed solution into the column. A 60 cm long & 6.5 cm inner dia acrylic cylinder is used as adsorption column. The use of acrylic limits the use of column for organic solutions to low and moderate concentrations while it is suitable for most inorganic solutions up to a wide range of concentrations. The constant flow of liquid from the storage tank to the column is maintained using a rotameter 0-5 lt/h capacity. In the adsorption studies on phosphate treated sawdust, separation of Zn 2+, Cu 2+, and Ni+2, their removal and recovery from electroplating wastes was done by Siddiqui et al.,(Siddique, Bilquees Ara, Sharma P.P. and Sultan Mohammad, 1998). Phosphate treated sawdust was found to exhibit remarkable adsorption capacity for Cu+2, Zn+2 and Ni+2. The effect of initial concentration in electroplating wastes waters; pH and adsorption doses were extensively studied. It was reported that the data Figure.1: Experimental set-up 8 Caledonian Journal of Engineering Volume04, Number 01, January - June 2008 4.0 EXPERIMENTAL PROCEDURE 4.4 Sample Analysis The steps involved in experimentation are: Solutions Required: • Preparation of coconut shell coke 1. • Preparation of feed solution 2. 3. 4. 5. 6. 7. • Experimentation • Sample analysis 4.1 Preparation of Coconut Shell Coke Coconut shells are made up of stone cells and are hard, porous and impregnated with lignin, tannins and contain a small amount of oil. The coconut shells were surface treated to remove all the fibers on its surface and then subjected to partial oxidation. Partial oxidation is necessary in order to prevent the formation of ash. The final material i.e. the black and brittle activated carbon is crushed in ball mill. The material is sieved and segregated according to the desired sizes (2.855 mm, 1.85 mm, and 1.55 mm). The adsorbent is then subjected to continuous washing for about 3 hours to remove the intense black color. 0.01N EDTA (Ethylene diamine tetra acetic acid) solution. 0.01N Zinc sulphate solution. Indicators. EBT (Erichrome Black – T ) FSBF (Fast Sulphon Black – F ) Concentrated Ammonia Solution. Buffer solution :- It was prepared by dissolving 17.5 gms of ammonium chloride in 142 ml of concentrated ammonia solution and the solution was made upto the mark in 250 ml volumetric flask. Copper ions present in solution can be determined by titrating with a solution of EDTA which was standardized against a standard solution of zinc sulphate. The desired pH for the titration was maintained by using concentrated ammonia solution. Fast Sulphon Black – F (FSBF) is used as indicator, gives wine red coloured complex with copper ions. After all the Cu+2 is complexed the indicator will set free and the end point will be indicated by colour change from wine red to dark green. The reactions are as follows:- 4.2 Preparation of Feed Solution Copper sulphate feed solution ‘A’ (70 ppm), ‘B’ (100 ppm) and ‘C’ (150 ppm) are prepared by dissolving 5.498 grms, 7.8552 grams and 11.78 grms respectively in 20 lts of distilled water. The solution was constantly stirred for about half an hour to obtain uniform concentration. Cu +2 + FSBF → Cu-FSBF (complex) Cu +2 +EDTA → Cu- EDTA (wine red) Cu-FSBF +EDTA → Cu-EDTA +FSBF (dark green) 4.3 Experimentation 5.0 RESULT & DISCUSSION The column is filled with the adsorbent and the storage tank with copper sulphate solution of desired concentration. The pump is switched on and the rotameter is adjusted to a desired flow rate. The solution is allowed to pass through the bed and sample coming out through the bed is collected. The same procedure is repeated for different initial concentrations, flow rates and for different adsorbent sizes. 5.1 Break through Curve The experiments are carried out for known values of initial concentration of metal ion, adsorbent size, and flow rate. The data on adsorption is plotted in Figure.2. The sudden rise in outlet concentration is observed immediately at a time of about 10 min, which may be considered as break through point. There after concentration in the effluent increased continuously with time up to t = 150 min and then remained constant. This indicates the saturation state of the bed. The portion of effluent concentration curve between t= 5 min to t = 150 min was termed the break through curve. Initial concentrations – 150ppm, 100ppm and 70 ppm Flow rate – 5 lph, 10 lph and 15 lph. Adsorbent sizes – 2.85 mm, 1.85 mm and 1.55mm. 9 Caledonian Journal of Engineering Volume04, Number 01, January - June 2008 90 80 Cout(ppm) 70 60 50 5 LPH 40 30 20 10 0 0 100 200 300 400 tim e (m in) Figure.2: Outlet concentration versus time (Initial Concentration = 100 ppm; Flow Rate = 5 LPH; and Size of Adsorbent = 1.85 mm). 1.85 mm and 140 min in case of 1.55 mm and thereafter the percentage of solute adsorbed remained nearly constant when the bed approached a state of equilibrium. With increase in particle size, the time taken for saturation of bed increases enormously which indicates that larger the size of the adsorbent particle, greater will be the interfacial area available for adsorption and the bed slowly approaches saturation. 5.2 Effect of Flow Rate To study the effect of flow rate on removal of copper ion by adsorption, other variables such as initial concentration of metal ion, and adsorbent size are kept constant during the period of experimentation. The data on adsorption are plotted in Figures 3 and 4. It was observed that the percent removal of metal ion decreased continuously with increase in time up to 180 min for 5 LPH, 90 min in the case of 10 LPH and 25 min in the case of 15 LPH and thereafter the percentage of solute adsorbed remained nearly constant when the bed approached a state of equilibrium. With increase in flow rate, the time taken for saturation of bed decreases enormously which indicates that more the mass of metal ions entering the bed, faster will be the saturation of bed. 5.4 Effect of Initial Concentration To study the effect of initial concentration on removal of copper ion by adsorption, other variables such as particle size, height of the adsorbent bed and flow rate are kept constant during the period of experimentation. The data on adsorption are plotted in Figure 7 and 8. It was observed that the percent removal of metal ion decreased continuously with increase in time up to 130 in case of 150 ppm, 180 min in case of 100 ppm and 260 min in case of 70 ppm and thereafter the percentage of solute adsorbed remained nearly constant when the bed approached a state of equilibrium. With increase in initial concentration, the time taken for saturation of bed decreases enormously which indicates that higher the initial concentration larger will be the driving forces. With increase in driving force the bed adsorbs much more quantity of metal ions and thus the time taken for saturation of bed decreases. 5.3 Effect of Particle Size To study the effect of particle size on removal of copper ion by adsorption, other variables such as initial concentration of metal ion, and flow rate are kept constant during the period of experimentation. The data on adsorption are plotted as shown in Figures 5 and 6. It was observed that the percent removal of metal ion decreased continuously with increase in time up to 180 min in case of 2.85 mm, 150 min in case of 10 Caledonian Journal of Engineering Volume04, Number 01, January - June 2008 120 Cout(ppm) 100 80 5 LPH 10 LPH 60 15 LPH 40 20 0 0 100 200 300 400 tim e (m in) Figure.3: Outlet concentration versus time for different flow rates (Initial Concentration = 100 ppm and Size of Adsorbent = 2.85 mm). 60 % Adsorption 50 40 5 LPH 30 10 LPH 15 LPH 20 10 0 0 100 200 300 400 tim e(m in) Figure.4: Percentage of Adsorption versus time for different Flow rates (Initial Concentration = 100 ppm and Size of Adsorbent = 2.85 mm). 11 Caledonian Journal of Engineering Volume04, Number 01, January - June 2008 100 Cout (ppm) 90 80 70 2.85 mm 60 50 40 1.85 mm 1.55 mm 30 20 10 0 0 100 200 300 400 time(min) Figure.5: Outlet concentration versus time for different particle sizes (Initial concentration = 100 ppm and Flow Rate = 5 LPH). % Adsorption 90 80 70 60 1.55 mm 50 1.85 mm 40 30 20 2.85 mm 10 0 0 100 200 300 400 time(min) Figure.6: Percentage of adsorption versus time for different particle sizes (Initial Concentration = 100 ppm and Flow Rate = 5 LPH). 12 Caledonian Journal of Engineering Volume04, Number 01, January - June 2008 140 120 Cout(ppm) 100 70 ppm 80 100 ppm 60 150 ppm 40 20 0 0 100 200 300 400 time(min) Figure.7: Outlet concentration versus time for different initial concentrations (Flow Rate = 5 LPH and Size of Adsorbent = 2.85mm). 70 % Adsorption 60 50 70 ppm 40 100 ppm 30 150 ppm 20 10 0 0 100 200 300 400 time(min) Figure.8: Percentage of adsorption versus time for different initial concentrations (Flow rate = 5 LPH and Size of Adsorbent = 2.85mm). 13 Caledonian Journal of Engineering Volume04, Number 01, January - June 2008 6.0 CONCLUSIONS The effects of various parameters like flow rate, particle size and initial concentration are studied and the following conclusions were drawn. • • the increase in flow rate and initial concentration decreases saturation time of the bed. the increase in particle size results in an increase in saturation time of bed. REFERENCES [1] Gupta., Vinod K., Ali., and Imran. (2000) Utilization of bagasse fly ash for the removal of copper and zinc from waste water, Sep Purif Technol 18(2), 130 – 140 [2] McCabe W.L., Smith J.C. and Peter Harriot, Unit Operations of Chemical Engineering, 5th Edition, McGraw Hill. [3] Perry and Chilton, Perrys Chemical Engineers Hand book, 7th Edition , McGraw Hill. [4] Siddique., Bilquees Ara., Sharma, P.P.; Sultan Mohammad., (1998) Adsorption studies on phosphate treated sawdust ; separation of Cr (vi), Zn+2,Ni+2 , Cu+2 and their removal and recovery from electroplating waste, Aurelia. Sci. Technol Environ. Prot, 3 (2), 103 – 110. [5] Seco Aurrora., Marzal Paula., Gabaldon Carmen., Ferrer Jose., (1999) Study of adsorption of cadmium and zinc on to an activated carbon. Influence of pH , cation concentration and adsorbent concentration, Sep. Sci. Technol. [6] Treybal R.E., Mass –Transfer Operations, 3rd Edition, McGraw Hill. 14 Caledonian Journal of Engineering Volume04, Number 01, January - June 2008 COAST DOWN FACTOR TO INVESTIGATE THE TRIBOLOGICAL BEHAVIOUR OF LUBRICANTS IN JOURNAL BEARING. R. Edwin Browne1, Dr. K. P. Ramachandran2, Dr. A.K.M. De Silva3, Prof. D.K. Harrison4 1&2 Department of Mechanical and Industrial engineering, Caledonian College of Engineering, Sultanate of Oman. 1. Corresponding Author, edwinbrowne@caledonian.edu.om, 2.ramkp@caledonian.edu.om. 3&4 School of Engineering and Computing, Glasgow Caledonian University, Cowcaddens Road, Glasgow, G4 0BA, Scotland, UK .3.A .DeSilva@gcal.ac.uk,4.D.K.Harrison@gcal.ac.uk ABSTRACT The run down time condition of any rotating system when analysed will lead to useful information for condition based maintenance. When the power supply is cut off to any rotor system, the total momentum gained during the sustained operation will dissipate and the system will come to rest. The time elapsed between the power supply cutoff and the system to stop is defined as Coast Down Time (CDT). The graphical representation of the speed with respect to CDT is known as Coast Down Time- Profile (CDT-P).This paper presents an experimental investigation conducted on the horizontal rotor system with full journal bearing at different cut-off speeds and tribological conditions. Investigation indicates that CDT is a dependant parameter upon inertia forces on the system components, mechanical and environmental conditions and tribological behaviour. In this paper an empirical relation for Coast Down Factor (CDF) has been developed for useful interpretation of the CDT-P in order to understand the rotating system. Observation reveals that CDF has potential benefits which could be used as a diagnostic parameter for condition monitoring in order to ascertain the tribological behaviour of lubricants in journal bearings. KEY WORDS: Coast Down Time, Coast Down Time Phenomenon, Deceleration, Coast Down Time - Profile, Condition monitoring, Diagnostic parameter. 1.0 INTRODUCTION Fluid film bearings seldom operate to the expected standard with inferior lubrication and performance. Continuous/sustained operation of journal bearing can lead to changes to good oil in film stiffness and damping characteristics, load carrying capacity, oil film temperature and stability of the rotor. Continuous monitoring of lubrication practice and wear would facilitate analysis on wear / lubricant degradation. It would also help to detect the oil contaminants as well as the deterioration of the lubricant. The main purpose of a good maintenance programme is to achieve optimum bearing life. Failure free running of any rotating equipment depends on the effective maintenance programme adopted for bearing. The selection of a reliable condition monitoring technique becomes important to properly diagnose and asses the performance of the lubrication and the deterioration of the lubricant for a journal bearing (Ramachandran et.al,1996). Tribomonitoring is considered to be more effective and reliable for rotating machines supported between bearings (Ran Barron 1996). The performance of the journal bearing is entirely dependent on the functions of the lubrication related to the basic properties of the 15 lubricant like viscosity, density, operating speed as well as the thermal properties like temperature and pressure (Ramachandran et.al, 1996 a). Any rotor system will come to a stop when the power supply is cut-off to the system. The momentum gained by the rotor system during the sustained operation will gradually dissipate until the system comes to a halt. This phenomenon during the deceleration period is known as Coast Down Phenomenon (CDP). The time elapsed from the moment the power supply is cut off until the system comes to rest is known as Coast Down Time (CDT) (Craig 1996 and Daughertyet.al,1976). Extensive investigations were conducted by Xistris et al 1974 on vertically supported motors by rolling element and established that CDT monitoring could be used as a potential health condition monitoring tool. CDT monitoring is simple as well as economical to implement. Ramachandran, 1992, Santhanakrishnan et.al, 1983 conducted experiments on a flexible rotor system with hydrodynamic journal bearings to investigate the influence of the lubricant on CDT and found that the deceleration speed vs. CDT resembled the Stribeck diagram of friction. Caledonian Journal of Engineering Volume04, Number 01, January - June 2008 Previous research had shown that optimization in relation to selection of the lubricant under a given condition could be achieved through CDT monitoring (Edwin Browne et. al,.2005). In this paper Coast Down Time- Profile (CDT-P) is first obtained to ascertain the tribological behaviour of lubricant inside a journal bearing by varying operating conditions, like different cut-off speeds, different lubricants and different lubricant line pressures. Furthermore Coast Down Factor (CDF) a dimensionless factor is deduced, using an empirical relation, which is developed to interpret the CDT-P. CDF is defined as a ratio of speeds measured at constant intervals of time during coast phase of any rotating system until the rotor system reaches a permanent stop after power supply is cut off to the rotating system, with respect to the initial speed drop at the first time interval. Since a marginal deviation is noticed among the CDT –P, arriving a diagnostic decision, based on the changes of CDT-P in the transition regions and on the trend of speed reduction during the deceleration period ,is found to be little difficult and complex . Moreover the CDF trend is expected to reflect a change magnified in order to understand the behaviour during deceleration of a rotor system. The present investigation is attempted to study CDF as a potential parameter that could be integrated into the condition monitoring programme. 2. EXPERIMENTAL SETUP order to position the journal at the centre of the bearing clearance, a pre load frame, consisting of a rolling element bearing with three support springs and set screws, was mounted on the base frame, very close to the rotor between the drive side support and rotor. The selected lubricant was circulated through the journal bearing by a gear pump. The rotor system was driven by a variable speed AC motor. Instrumentation with an update time of 195 millisecond with built in RS-232 PC interface was used for recording time. Bal Pac 1200 vibration data collector was used for measuring the vibration at a steady running state. The critical speed for the set up was computed approximately at 6000 rpm. Experiments were conducted for different cut-off speeds, different lubrication and a range of lubrication line pressures to analyse the tribological behaviour of the lubricants. Different cut-off speeds like 1000 rpm, 2000 rpm, 3000 rpm, 4000 rpm and 5000 rpm were selected to run the rotor system along with selected lubricants, under different values of lubricant line pressure for a considerable amount of time, to attain sufficient momentum. Four different lubricants, good oils SAE 90, SAE 40, used oils SAE 90 and SAE 40 were used for experiments with three different lubricant line pressures 80 kPa, 100 kPa and 120 kPa. During the steady state operation of the rotor system, the vibration data was recorded in both vertical and horizontal directions from both the drive end bearing and journal end bearing. Power to the rotor system was made to trip at the selected cut-off speed and the rate of change of speed was recorded using the instrumentation during the deceleration period until the rotor comes to a halt. Properties of lubricants: SAE 90 *SAE 90 SAE 40 **SAE 40 Good Used Good Used Oil Oil Oil Oil Viscosity @40oC - 161.3 140.9 151.3 110.9 in mm2/s Density @15oC - 0.89944 0.8758 0.8989 0.8965 in g/cm3 Water content Nil Nil Nil Nil Sulphur- 0.84 1.24 0.99 1.28 % wt *SAE 90 used oil: Experimental set-up shown in figure 1 was used for the investigation. Sample of oil collected from a power transmission gear box of an automobile vehicle which had completed one useful life cycle time of running 45,000 km. Figure.1: Experimental set-up. The rotor system consisting of a rotor weighing 800 grams was mounted at the centre of a 10 mm shaft and supported by bearings. The drive end of the shaft was supported by an anti-friction bearing and the non drive end was supported by a full journal bearing. In 16 ** SAE 40 used oil: Sample of oil collected from automobile engine lubrication which had completed one useful life cycle time of running 5000 km. Caledonian Journal of Engineering Volume04, Number 01, January - June 2008 Specification of the journal bearing: Length of the journal = 25.4 mm Diameter of the journal = 25.06 mm Diameter of the journal bearing = 25.38 mm Clearance of the bearing = 0.32 mm SAE 4 0 Go o d Oil - 8 0 kPa SAE 4 0 Go o d Oil - 10 0 kPa SAE 4 0 Go o d Oil - 12 0 kPa SAE 4 0 Us ed Oil - 8 0 kPa SAE 4 0 Us ed Oil - 10 0 kPa SAE 4 0 Us ed Oil - 12 0 kPa 2500 Speed 2000 3. RESULT AND DISCUSSIONS 1500 1000 500 Figure 2 depicts the typical characteristics of a CDT-P expected of a journal bearing to emphasize that the CDT-P is just the reverse of Stribeck frictional curve. The shape will be correlated to the frictional characters described by Raimondi & Boyd design curve. 0 0 390 780 1170 1560 1950 2340 2730 CDT in milliseconds F ric tio n Figure.4: CDT-P: SAE 40 Oil- 2000 RPM I II SAE 9 0 Go o d Oil - 8 0 kPa SAE 9 0 Go o d Oil - 10 0 kPa SAE 9 0 Go o d Oil - 12 0 kPa SAE 9 0 Us ed Oil - 8 0 kPa SAE 9 0 Us ed Oil - 10 0 kPa SAE 9 0 Us ed Oil - 12 0 kPa 3500 III 3000 Speed 2500 Speed I Hydrodynamic Lubrication Zone II Elastohydrodynamic or Mixed Lubrication Zone III Boundry Lubrication Zone I Hydrodynamic Lubrication Zone 2000 1500 1000 500 II Elastohydrodynamic or Mixed Lubrication Zone 0 0 585 III Boundary Lubrication Zone Figure.2: Friction Vs Speed SAE40 Good Oil - 80 kPa SAE40 Good Oil - 100 kPa SAE40 Good Oil - 120 kPa SAE40 Used Oil - 80 kPa SAE40 Used Oil - 100 kPa SAE40 Used Oil - 120 kPa 1000 Speed 800 Figure 3: CDT-P: SAE 40 Oil – 1000 RPM 600 400 200 390 780 1170 SAE 4 0 Go o d Oil - 8 0 kPa SAE 4 0 Go o d Oil - 10 0 kPa SAE 4 0 Go o d Oil - 12 0 kPa SAE 4 0 Us ed Oil - 8 0 kPa SAE 4 0 Us ed Oil - 10 0 kPa SAE 4 0 Us ed Oil - 12 0 kPa 4500 4000 3500 3000 2500 2000 1500 1000 500 0 0 585 1170 1755 2340 2925 3510 4095 4680 CDT in milliseconds Figure.6: CDT-P: SAE 40 Oil-4000 RPM SAE 90 good oil at 80kPa was found exhibiting a distinctive performance as expected from a good lubricant. The CDT-P obtained for SAE 90 good oil at 80kPa was very close to Figure 1. The generic profile of the coast down behaviour exemplifies the typical frictional characteristics of journal bearing of the rotor system under investigation in (Edwin Browne et.al, 2005). CDT-P for SAE 90 good oil at 80kPa has 0 0 Speed Figure.5: CDT-P: SAE 40 Oil -3000 RPM Observing the data, which were obtained during the steady state of running, it is evident that the CDT is increased with the increase in cut-off speed. The CDT-P for, SAE 40, good oils and used oils for all the tested cut-off speeds with different lubricant line pressures are given in Figures 3 - 7. 1200 1170 1755 2340 2925 3510 CDT in milliseconds 1560 CDT in milliseconds Figure.3: CDT-P: SAE 40 Oil – 1000 RPM 17 Volume04, Number 01, January - June 2008 shown a distinct and smooth curve in both the hydrodynamic lubrication and boundary lubrication zones, which has very closely followed the inverse of Stribeck curve. CDT was found increasing with the increase of cut-off speeds (Edwin Browne et al11). 35 E Caledonian Journal of Engineering 10 5 0 SAE 90 oil , used at all tested pressures and cut-off speeds, exhibited CDT-P with lot of deviation throughout the deceleration period, when compared with that of CDT-P of good oil. The degradation of the used oil decreases the CDT and increases the vibration amplitude (Edwin Browne et al11). 0 SAE 4 0 Go o d Oil - 8 0 kPa SAE 4 0 Go o d Oil - 10 0 kPa SAE 4 0 Go o d Oil - 12 0 kPa SAE 4 0 Us ed Oil - 8 0 kPa SAE 4 0 Us ed Oil - 10 0 kPa SAE 4 0 Us ed Oil - 12 0 kPa 5000 Speed 4000 3000 2000 1000 1170 S AE 90 Go o d Oil - 80 kP a S AE 90 Go o d Oil - 100 kP a S AE 90 Go o d Oil - 120 kP a S AE 90 Us e d Oil - 80 kP a S AE 90 Us e d Oil - 100 kP a S AE 90 Us e d Oil - 120 kP a 60 40 20 0 0 585 1170 1755 CDT 2340 E Figure.9: CDF- SAE 90 Oil -2000RPM 80 70 60 50 40 30 20 10 0 S AE 90 Go o d Oil - 80 kP a S AE 90 Go o d Oil - 100 kP a S AE 90 Go o d Oil - 120 kP a S AE 90 Us e d Oil - 80 kP a S AE 90 Us e d Oil - 100 kP a S AE 90 Us e d Oil - 120 kP a 0 780 1560 2340 3120 CDT Figure.10: CDF-SAE 90 Oil-3000 RPM S AE 90 Go o d Oil - 80 kP a S AE 90 Go o d Oil - 100 kP a S AE 90 Go o d Oil - 120 kP a S AE 90 Us e d Oil - 80 kP a S AE 90 Us e d Oil - 100 kP a S AE 90 Us e d Oil - 120 kP a 80 60 E 6000 780 CDT 80 The degradation of the lubricant decreases the CDT and increases the vibration level which complements the earlier findings (Edwin Browne et. al, 2006). As there were little deviations noticed among the CDT-P, a distinct conclusion based CDT-P analysis was not possible. In order to study the CDT-P in a meaningful way as one of the condition monitoring parameters for a rotor system, a dimensionless factor has been defined as Coast Down Factor (CDF) denoted with a letter ‘E’ as Edwin’s factor, which can be calculated using the following empirical formula. En = {Cs – Nn) / (Cs-N1)} En - Coast Down Factor at Tn Cs - Cut-off speed N1 – Rotor speed after first interval milliseconds (T1). Nn – Rotor speed measured at successive intervals in milliseconds (Tn). 390 Figure.8: CDF-SAE 90 Oil-1000 RPM E Less viscosity and variations in the bulk properties make the CDT-P or SAE 40 oil, different from the expected form. CDT-P confirms that SAE 40 oil is not a suitable lubricant for the selected journal bearing under investigation. SAE 40 good oil at 100kPa was found producing CDT-P very close to the expected one but the vibration amplitude was found to be more whereas the CDT was less. S AE 90 Go o d Oil - 80 kP a S AE 90 Go o d Oil - 100 kP a S AE 90 Go o d Oil - 120 kP a S AE 90 Us e d Oil - 80 kP a S AE 90 Us e d Oil - 100 kP a S AE 90 Us e d Oil - 120 kP a 30 25 20 15 40 20 0 0 0 780 1560 2340 3120 3900 4680 5460 6240 CDT in milliseconds 0 975 1950 2925 3900 CDT Figure.11: CDF- SAE 90 Oil-4000 RPM Figure.7: CDT-P: SAE 40 Oil -5000RPM 18 Caledonian Journal of Engineering SAE 90 Go o d Oil - 80 kP a SAE 90 Go o d Oil - 100 kP a SAE 90 Go o d Oil - 120 kP a SAE 90 Us ed Oil - 80 kP a SAE 90 Us ed Oil - 100 kP a SAE 90 Us ed Oil - 120 kP a 200 150 E Volume04, Number 01, January - June 2008 100 50 0 0 1170 2340 3510 4680 5850 CDT complements the author earlier findings with validation from vibration data (Edwin Browne et al10). The changes in the bulk properties and low viscous for SAE 40 good oil are the reasons for getting a different CDF trend as well as different CDF values for different cut-off speeds. Comparatively low CDF values confirm that the SAE 40 oil is not a suitable lubricant for the rotor system under investigation. Among the graphs obtained for CDT-P and CDF, it is clearly illustrated that the CDF graphs are better profiled to indicate the performance of lubricants distinctly, during deceleration. The functional characteristics of journal bearing under various regimes of lubrication are fully understood by CDF for rotor system under investigation. Figure.12: CDF-SAE 90 Oil -5000 RPM 4. CONCLUSION Figure 8-12 the trend curves plotted for E and CDT values to analyse and interpret the CDT-P of the rotor system tested with SAE 90 oil for different speeds and different lubricant line pressures. Figure 13-17 are the trend curves plotted for E and CDT values to analyse and interpret the CDT-P of the rotor system tested with SAE 40 oil for different speeds and different lubricant line pressures. It was observed that the trend plot of CDF for both SAE 90 and 40 good oils at 80kPa were distinctively positioned well above the other curves with different line pressures and at different conditions. This indicates that there exists a potential function related to CDF when analysed between different operating conditions for a rotor system. The steady state operation of the rotor system allows the journal bearing to operate within the region of hydrodynamic lubrication. The lesser speed difference for the SAE 90 good oil at 80kPa, at the first interval as soon as the power to the rotor was tripped, signifies that the rotor was subjected only to the fluid frictional resistance. As the speed reduced and when the rotor entered the mixed lubrication zone the CDF value was found increasing. The CDF was found reaching maximum when the rotor came to rest following through the boundary lubrication zone. It was observed that when using different lubricants or different lubricant line pressures, the fluid resistance in the hydrodynamic lubrication zone was getting changed, the effect of which had been noticed in the form of dominant speed reduction during the first interval. This signifies the behaviour of the lubricants in the journal bearing. It was observed that SAE 90 good oil at 80kPa was showing an improved performance exhibiting the typical characteristics of the good lubricant. CDF obtained for SAE 90 oil confirms that the best suited lubricant for the tested rotor system is SAE 90 oil at 80kPa, which 19 In the present investigation, the tribological behaviour of lubricants in a rotor system was studied under different operating conditions. Results are in agreement with earlier findings that CDT increases with an increase in cut-off speed. The degradation of the lubricant decreases the CDT and increases the vibration level which complements the earlier findings (Edwin Browne et al11, 12). This paper presented the usefulness of CDF properties to overcome the perceived difficulties encountered with CDT-P in demarking the transition regime during deceleration. The CDF values indicate the characteristic behaviour of the lubricants when tested with different operating conditions. It was found that the trend curves and analysis of CDF, determined at equal intervals of time during deceleration of the rotor system until the rotor stopped, convey potential information for diagnosing the condition of lubricants in a rotor system which would be useful for proactive maintenance. REFERENCES [1] Craig.R.J (1976) Application of coast down monitoring techniques for vertical shaft motors, DTNSRDC Rep, 76, p.13. [2] Daugherty.T.L., Craig. R.J., January (1976) Coast down time as a mechanical condition indicator, DTNSRDC Rep, 4547. [3] Edwin Browne. R., De Silva.A.K.M., Ramachandran.K.P. Harrison.D.K. Sharif.M.EL (2005) An evaluation of bearing lubrication and the selection of lubricants using CDT analysis as a condition monitoring parameter, Regional conference in Recent trends in Maintenance Management, Sultanate of Oman, Paper no 4, p.25-29. Caledonian Journal of Engineering Volume04, Number 01, January - June 2008 [4] Edwin Browne. R., De Silva.A.K.M., Ramachandran.K.P., Harrison.D.K. (2006) CDT analysis is used as a condition monitoring parameter to study the Tribological behaviour of bearing lubricant under different operating pressure, Porc. ESDA 2006, 8th Biennial ASME conference, Torino, Italy, ISBN: 0-7918-377903, ESDA 2006-95021, p.111. [5] Edwin Browne. R., De Silva.A.K.M., Ramachandran.K.P., Harrison.D.K. (2006) Evaluation of tribological behaviour in journal bearing using Coast Down Time analysis, LUBMAT 2006, European conference in lubrication management and technology, Preston, UK, ISBN 1-901922-58-8, p.9. [6] Ramachandran.K.P. and Ramakrishna.A. (1996) Oil analysis for failure prevention of plant machinery, National seminar on failure analysis, HIMER, p.8893. [7] Ramachandran.K.P. and Ramakrishna.A. (1996) Lubrication scheduling in manufacturing industries and its optimisation, Transaction of Industrial product finder, p.244 -248. [8] Ramachandran K.P. (1992) Coast down time analysis as a condition monitoring tool, Indian Journal of Maintenance, National Productive Council, 11, p.45-46 [9] Ran Barron (1996) Engineering Condition Monitoring Practices, Methods and Application (Longman, UK) [10] Santhanakrishnan.G., Prabhu.B.S. Rao. B.V.A. (1983) An investigation of tribological effects on coast down phenomenon in horizontal machinery, Journal of wear, 91, p25-31. [11] Xistris.G.D, Watson.D.C (1974) Proc. “Mechanical Failures Prevention” Group, Gaithersbarg, MD, National Bureau of Standards, Washington, DC. [12] Xistris G.D., Watson.D.C (1975) ASME Prepr., 75-DE-6, New York. 20 Caledonian Journal of Engineering Volume04, Number 01, January - June 2008 PROBALISTIC ANALYSIS OF A SYSTEM WITH TWO TYPES OF REPAIRMAN AND PATIENCE TIME WHEREIN THE INITIAL REPAIR IS UNDERTAKEN BY THE ORDINARY REPAIRMAN Vandna Bhagat Dept. of Humanities & Applied Science Advanced Institute of Technology & Management Vill. Aurangabad, Distt. - Palwal (121102), Haryana, India ABSTRACT Two-unit cold standby system with two types of repairman is studied. The failed unit is first undertaken by an ordinary repairman for repair who may not be able to do complex repairs. Idea of patience time, i.e., the maximum time of waiting for an expert while the ordinary repairman is trying to repair the failed unit, is also introduced. The expert repairman may or may not be found available when required. This model has been compared with the two models wherein the expert is called first to do the repair, if available; otherwise ordinary repairman is called who if shows inability of repairing the failed unit, the expert is called. unit waits for repair till the availability of the expert repairman whereas in Model 2, the expert repairman is made available immediately by paying some additional amount. However, the ordinary repairman may take a lot of time to declare himself unable to complete the repair successfully. Thus, we should not wait till such declaration after a very long time. We should wait up to some limited amount of time called as patience time. 1.0 INTRODUCTION A large literature exists in the areas of reliability of standby systems under the assumption that the perfect repairs are done for the failed units. Some researchers such as (A. Kumar, S.K.Gupta and G.Taneja, 1996 ), (R.K.Tuteja and G.Taneja, 1992) (R.K.Tuteja, G.Taneja and A.Malik, 2000-2001) (R.K.Tuteja, R.T.Arora and Gulshan Taneja, 1991) (S.K. Singh and R.P. Singh,1989), and (V.Goyal and K.Murari, 1984) considered the two types of repair, i.e., by an ordinary repairman and by an expert with assumption that ordinary repairman may not be able to do some complex repairs and then expert is called. They considered immediate availability of the expert whenever required. However, the expert may not always be available whenever required. Rizwan (S.M.Rizwan, 2007) discussed reliability analysis of two unit system with two repairmen wherein ordinary repairman may not be available on requirement but the expert is available on requirement. Taneja et.al (G.Taneja, V.Naveen and D.K.Madan, 2001) analysed reliability and profit analysis of a system with an ordinary and an expert repairman wherein the latter may not always be available (G.Taneja, V.Naveen and D.K.Madan, 2001) and also assumed that on failure of a unit, an expert repairman is called first to do the repair, if available. If not available, then ordinary repairman is called who may not be able to do some complex repairs. They discussed two models. If the ordinary repairman finds himself unable to repair the unit, then in Model 1, the failed We in the present paper, introduce the concept of such patience time while dealing with a two-unit cold standby system with the assumption that the failed unit is first undertaken by the ordinary repairman who may not be able to do some complex repairs. If the ordinary repairman is unable to repair the failed unit, an expert is called if he is available. If he is not available, we wait till the expert becomes available or both units get failed or the patience time is completed whichever is earlier. Here, the patience time is maximum time of waiting for the expert repairman. It is assumed that the expert repairs all the units which fail during his stay at the system. The model is analysed stochastically by making use of semi-Markov processes and regenerative point technique and the expressions for various measures of system effectiveness such as mean time to system failure, steady state availability, total fraction of busy time of ordinary repairman, total fraction of busy time of expert repairman on ordinary/special visit. Expected number of visits by an ordinary and an expert repairman are found out. Profit is also calculated using the above measures. Graphs are 21 Caledonian Journal of Engineering Volume04, Number 01, January - June 2008 plotted for particular case .The model is also compared with the models discussed by G.Taneja, V.Naveen and D.K.Madan (G.Taneja, V.Naveen and D.K.Madan, 2001). Bi(t), (Bi(e)(t)) NOTATIONS Bsi(t) O cs λ p q a1 operative cold standby constant failure rate of the operative unit probability that the expert repairman is available probability that expert repairman is not available probability that ordinary Wi(t) Vi(t),(Vi(e)(t)) repairman is able to repair the failed unit a2 probability that ordinary through any other regenerative state. probability that ordinary (expert) repairman is busy at instant ‘t’ given that the system started from regenerative state i at t = 0 probability that expert repairman on his special visit is busy at instant t given that the system started from regenerative state i at t=0 probability that a repairman is busy with the system initially in regenerative state i is busy at time t without passing through any other regenerative state expected number of visits of an ordinary (expert) repairman in (0,t] give that the system from regenerative state i at t = 0 2.0 TRANSITION PROBABILITIES AND MEAN SOJOURN TIMES repairman is unable to repair the failed unit g1(t), G1(t) The transition diagram showing various states of transition of the system is shown in Figure.1. The epochs of entries into the states 0, 1, 2, 3, 5, 6 and 8 are regenerative points and thus these are regenerative states. States 4, 5, 7, 8 and 9 are down states. The non-zero elements pij are given by: p.d.f. and c.d.f. of time to repair by ordinary repairman g2(t), G2(t) p.d.f. and c.d.f. of time to repair by expert repairman h(t), H(t) w(t), W(t) Fr Fre Fres FR FRe FRes φi(t) p.d.f. and c.d.f. of patience time p.d.f. and c.d.f. of waiting time failed unit under repair of ordinary repairman failed unit under repair of expert repairman failed unit under special repair of the expert repair is continuing from previous state by the ordinary repairman repair is continuing from previous state by the expert repairman special repair by the expert is continuing from previous state. c.d.f of first passage time from p01 = 1, p10= a1g1*(λ), p12 = pa2g1*(λ), p13 = qa2g1*(λ), p14 = 1−g1*(λ), p11(4) = a1[1−g1*(λ)], p15(4) = a2q[1−g1*(λ)], p18(4) = a2p[1−g1*(λ)], p20 = g2* (λ), p27 = 1−g2*(λ), p22(7) = 1−g2*(λ), p32 =E1*(λ), p35 = λ E2*(λ), p36 = E3*(λ) ,p56 =1, p60 = g2*(λ), p69 = 1−g2*(λ), p66(9) = 1−g2*(λ), p82= 1 By these transition probabilities, it can be verified that p10 + p12 + p13 + p14 = p10 + p12 + p13 + p11(4) + p15(4) + p18(4) = 1, regenerative i to failed state Ai(t) probability that the system is up at p20 + p27 = p20 + p22(7) = 1, p32 + p35 + p36 = 1, p56 = 1, p60 + p69 = p60 + p66(9) = 1, p82 = 1 instant t given that system started from regenerative state i at t = 0 Mi(t) Also µi, the mean sojourn time in state i are probability that system is up initially in regenerative state i is up at time t without passing 22 Caledonian Journal of Engineering Volume04, Number 01, January - June 2008 + q13(t) © A3(t) + q15(4) (t) © A5(t) + q18(4) © A8(t) * * µ0 = 1 , µ1 = 1 − g1 (λ ) , µ2 = 1 − g 2 (λ ) , µ3 = E1* λ λ λ * 1 − g ( λ ) (λ), µ6 = 2 A2(t) λ = M3(t) + q32(t) © A2(t) + q35(t) © A5(t) + A3(t) q36(t) © A6(t) The unconditional mean time taken by the system to transition for any regenerative state j, when it (time) is counted from epoch of entrance into that state i is mathematically stated as : ∞ mij = ∫ t dQij (t) = − qij* ′(0) 0 ∫ G1(t)dt=K1(say) A6(t) A8(t) = M6(t) + q60(t) © A0(t) + q66(9)(t) © A6(t) = q82(t) © A2(t) M2(t)= The steady state availability of the system is given by A0 = lim s A0*(s) = N1/D1 0 ∞ ∫ G 2 ( t )dt = K 2 (say) 0 m60 + m69 = µ2, m60 + m66(9) = K2 s →0 ∞ where N1 = µ0 p20 p60 (1−p11(4)) +µ1 p20 p60 + µ2[p60 (p18(4) + p12+ p32) + p20 p13 (1−p32) + p15(4) p20] + µ3 p13 p20 p60 and D1 = µ0 p18(4) p20 p60 + K1 p20 p60 + K2 [p60 (p12+p18(4)+p13 p32) + p20 (p15(4) + p13 − p13 p32) + p20 p60 (p13 p35 + p15(4))] + k3 p13 p20 p60 ∫ t [e−λt⎯E1(t) + λe−λtE2(t) + e−λt 0 E3(t)] dt = K3 (say) 3.0 MEAN TIME TO SYSTEM FAILURE Using probability arguments and recursive elations for Bie(t), BSie(t), Bi(t), Vie(t), VSie(t) and Vi(t), we obtain the following measures in steady-state: φ0(t) = Q01 (t)(s) φ1(t) φ1(t) = Q10 (t)(s) φ0(t) + Q12(t)(s) φ2(t) + Q13(t)(s) φ3(t)+Q14 φ2(t) = Q20 (t)(s) φ0(t) + Q27(t) φ3(t) = Q32 (t)(s) φ2(t) + Q35(t) + Q36(t)(s) φ6(t) φ6(t) = Q60 (t)(s) φ0(t) + Q69(t) The total fraction of the time for which the expert repairman is busy in his ordinary visit (B0e) = N2/D1 The total fraction of time for which the expert repairman is busy on special visits (BS0e) = N3/D1 Now, the mean time to system failure (MTSF) when the system starts from the state 0, is T0 = = q56(t) © A6(t) M3(t)= e−λt⎯H(t)⎯W(t) ; M6(t) = e−λt⎯G2(t) ∞ m 32+ m35 + m36 = A5(t) Where M0(t)= e−λt; M1(t)= e−λt⎯G1(t) ; e−λt⎯G2(t); Thus, m01 = µ0, m10 + m12 + m13 + m14 = µ1, m10 + m12 + m13 + m11(4) + m15(4) + m18(4) = m20 + m27 = µ2, m20 + m22(7) = = M2(t) + q20(t) © A0(t) + q22(7)(t) © A2(t) The total fraction of time for which the ordinary repairman is busy (B0) = N4/D1 ** 1 − φ 0 (s ) N = lim s →0 s D The expected number of visits per unit time by the expert repairman (V0e) = N5/D1 where N = µ0 + p01µ1 + µ2 (p01 p12 + p01 p13 p32) + K3 p01p13 + µ2 p01 p13 p36 and D = 1 − p10 − p12 p20 − p13 p36 p60 − p13 p20 p32 The number of special visits per unit time by the expert repairman (VS0e) = N6/D1 The number of visits by ordinary repairman per unit time (V0) = N7/D1 Where N2 = K2 p60 [p13 p32 + p12 + p18(4) p22(7)] N3 = p20 K2 [p60 (p13 p35 + p15(4)) + p15(4) + p13 (1−p32)] N4 = K1 p01 p20 p60 4.0 AVAILABILLTY ANALYSIS A0(t) = M0(t) + q01(t) © A1(t) = M1(t) + q10(t) © A0(t) + q11(4) (t) © A1(t) A1(t) + q12 (t) © A2(t) 23 Caledonian Journal of Engineering Volume04, Number 01, January - June 2008 N5 = p20 p60 [p12 + p18(4) + p13 p32]. N6 = p20[p13{p35 + p66(9)} + p15(4) p60 + p66(9)] N7 = (1−p11(4)) p20 p60 (ii) and D1 is already specified. 5.0 PROFIT ANALYSIS The expected total profit incurred to the system in steady state is given by P = C0 A0 − C1 B0 -C2 B0e − C3(V0e+ VS0e) − C4 V0 − C5 BS0e (iii) Where C0 = revenue per unit up time of the system. C1 = cost per unit time for which the ordinary repairman is engaged in repairing the failed unit. C2 = cost per unit time for which expert repairman is busy (after his ordinary visit) in repairing the failed unit. C3 = cost per ordinary visit and special visits by the expert repairman C4 = cost per visit by the ordinary repairman. C5 = cost per unit time for which expert repairman is busy (after his special visit) in repairing the failed unit It is, therefore, concluded that if cost per unit time for which the expert is busy, is such that P-P1>0 then we should call the ordinary repairman first and if P-P1<0 the expert repairman should be called first to repair the failed unit. 8.0 COMPARITIVE STUDY OF THE PRESENT MODEL WITH THE MODEL 2 DISCUSSED IN [2] From Figure.3 it is clear that the difference of profits (P-P2) decreases on increasing the values of cost (C5). Following conclusions can be drawn: 6.0 GRAPHICAL INTERPRETATION (i) For the graphical representation, the following particular case is considered g1(t) = α1 e−α1t , g2(t) = α2 e−α2t , h(t) = β1 e−β1t , w(t) =β2 e−β2t The behaviour of the MTSF and the profit w.r.t. failure rate (λ) for different values of repair rate (α1) have been studied through graphs wherefrom it has been seen that as failure rate increases, MTSF as well as profit decreases. However, their values become higher for higher values of repair rate (α1). (ii) (iii) 7.0 COMPARITIVE STUDY OF THE PRESENT MODEL WITH THE MODEL 1 DISCUSSED IN [2] Figure.2 shows the behaviour of difference of profits (P-P1) w.r.t. cost (C5) for different values of patience rate (β1). It is clear from the graph that (i) model of G.Taneja, V.Naveen and D.K.Madan is better or worse than the present model. If β1 = 15 and other parameters are fixed then these two models are equally good if C5 = 7556. If C5 > 7556 or < 7556 then first model of G.Taneja, V.Naveen and D.K.Madan is better or worse than the present model. If β1 = 25 and other parameters are fixed then these two models are equally good if C5 = 8000. If C5 > 8000 or < 8000 then first model of G.Taneja, V.Naveen and D.K.Madan is better or worse than the present model. If β1 = 5 and other parameters are fixed then these two models are equally good if C5 = 6627. If C5 > 6627 or < 6627 then second model of G.Taneja, V.Naveen and D.K.Madan is better or worse than the present model. If β1 = 15 and other parameters are fixed then these two models are equally good if C5 = 7341. If C5 > 7341 or < 7341 then second model of G.Taneja, V.Naveen and D.K.Madan is better or worse than the present model. If β1 = 25 and other parameters are fixed then these two models are equally good if C5 = 7750. If C5 > 7750 or < 7750 then second model of G.Taneja, V.Naveen and D.K.Madan is better or worse than the present model. It is, therefore, concluded that if cost per unit time for which the expert is busy, is such that P-P1>0 then we should call the ordinary repairman first and if P-P1<0 the expert repairman should be called first to repair the failed unit and even by paying some extra amount if not available. If β1 = 5 and other parameters are fixed then these two models are equally good if C5 = 6835. If C5 > 6835 or < 6835 then first 24 Caledonian Journal of Engineering Volume04, Number 01, January - June 2008 types of repairman, Journal of Decision and Mathematical Sciences, 5-6, p.59-74. REFERENCES [1] A. Kumar, S.K.Gupta and G.Taneja (1996) Comparative study of the profit of a two server system including patience time and instruction time, Microelectron.Reliab., 36(10), p.1595-1601. [5] R.K.Tuteja, R.T.Arora and Gulshan Taneja (1991) Stochastic behaviour of a two unit system with two types of repairman and subject to random inspection, Microelectron.Reliab., 31(1), p.79-83. [2] G.Taneja, V.Naveen and D.K.Madan (2001) Reliability and profit analysis of a system with an ordinary and an expert repairman wherein the latter may not always be available, Pure and Applied Mathematika Sciences, LIV (1-2), p.11-25. [6] S.K. Singh and R.P. Singh (1989)) Stochastic analysis of complex system with two types of repair facility and patience time for repair, International Journal of Management and systems, 5, p.143-156. [7] S.M.Rizwan (2007), Reliability Analysis of a two unit system with two repairman, Caledonian Journal of Engineering, .3, p.1-5. [3] R.K.Tuteja and G.Taneja (1992) Cost benefit analyses of two server, two unit, warm standby system with different types of failure; Microelectron.Reliab., 32, p.1353-1359. [8] V.Goyal and K.Murari (1984) Cost analysis of a two-unit standby system with two types of repairman, Microelectron. Reliab, 24, p.849-855 [4] R.K.Tuteja and G.Taneja and A.Malik (20002001) Reliability and profit analysis of a two-unit cold standby system with partial failure and two 25 Caledonian Journal of Engineering Volume04, Number 01, January - June 2008 26 Caledonian Journal of Engineering Volume04, Number 01, January - June 2008 STABILITY ANALYSIS OF MULTI DIMENSIONAL DISCRETE POLYNOMIAL EMPLOYING EQUIVALENT ONE-DIMENSIONAL POLYNOMIAL Sivanandam S.N. and Rajan.S Department of Computer Science and Engineering PSG College of Technology, Peelamedu, Coimbatore – 641 004, India Email : sns@mail.psgtech.ac.in , sr_tce@rediffmail.com ABSTRACT In this paper, the stability analysis of multi dimensional discrete polynomial is carried out with the help of its equivalent one-dimensional polynomial along with the suggested necessary conditions and Marden table. The proposed algebraic procedure is simple and straight forward in application and illustrated through examples. KEYWORDS: Stability, Multi dimensional, One dimensional, Marden table, Necessary condition.; H ( z 1 , z 2 ,..., z k ) = 1.0 INTRODUCTION A( z 1 , z 2 ,..., z p ) B( z 1 , z 2 ,..., z q ) ……………………………………..…………...1 where A(z1,z2,…,zp) and B(z1,z2,…,zq) are real mutually prime polynomials. The stability is the main desirable feature of all kinds of systems and in designing a given system, it is important to choose the system parameters so as to avoid the possible occurrence of unstable condition. The stability problem of multidimensional polynomials is receiving more attention in recent years in view of the emerging widespread applications. The multi dimensional discrete polynomials generally occur in the fields of image processing, geophysics and in processing of bio-medical, physical, sonar and radar data (Tzafestas, 1986). Other applications arise in obtaining realizability properties of impedances of networks and transmission lines which represent multidimensional systems. For stability investigation, Anderson and Jury (Jury, 1974), gave the conditions for the denominator polynomial B(z1,z2,…,zq) to be non q zero in the region I z i ≤ 1. In general, the i =1 stability condition is that, B( z 1 , z 2 ,..., z q ) ≠ 0 , for all q I i =1 zi ≤ 1 …………………………….....…………..2 Equation (2) can be restated as (q-1) necessary conditions and one sufficient condition as given from equations (3) and (4) respectively: The stability investigation of multidimensional discrete polynomial is very interesting and many schemes have been reported in (Strintzis, 1977, Jury, 1974, Bose, 1974 & 1979, Zaheb, 1980, 1982 & 1984, Tzafestas, 1988, Plotkin, 1985, Bauer, 1991, 1992 & 1994), each scheme has its own merits and applications. In this paper, a simple algebraic procedure is presented to test the stability of multidimensional systems. Necessary conditions: B( z 1 ,0 ,0 ,...,0 ) ≠ 0 B( z 1 , z 2 ,0 ,...,0 ) ≠ 0 z1 ≤ 1 {z1 } { } = 1 ∩ z2 ≤ 1 . . . 2.0 STABILITY THEOREM OF MULTI DIMENSIONAL SYSTEM B( z 1 , z 2 ,..., z q −1 ,0 ) ≠ 0 ⎫ ⎧q −1 ⎨ I z i = 1⎬ I i = 1 ⎭ ⎩ {z q− A linear time invariant multidimensional discrete system (Jury, 1974) can be represented by a multivariable transfer function of the form, 1 } ≤1 …………………………………....…………….3 27 Caledonian Journal of Engineering Volume04, Number 01, January - June 2008 i) At z = 1, F(1)>0 ii) At z = -1, F(-1)>0 (for neven)……………………………..6 F(-1)<0 (for n-odd) iii) |a0|<|an| Sufficient condition: ⎧ q −1 ⎫ ⎨ i I=1 z i = 1⎬ I ⎩ ⎭ B( z 1 , z 2 ,..., z q ) ≠ 0 {z q } ≤1 ……………….…………………………………4 If the conditions stated in equation (6) are satisfied, then Marden table (Marden, 1966) is used for testing Fn(z) which is given in Table 1. Thus the problem of testing the stability condition in equation (2) for a q-variable polynomial is equivalent to the problem of testing the necessary conditions (3) and the sufficient condition given in equation (4). Table 1: Marden table for Fn(z) in equation (5) In this paper, a simple algebraic scheme is proposed to test the equations given in equation (3) as well as the equation (4). This scheme is straight forward and easy to apply compared to other schemes given in (Strintzis, 1977, Jury, 1974, Bose, 1974 & 1979, Zaheb, 1980, 1982 & 1984, Tzafestas, 1988, Plotkin, 1985, Bauer, 1991, 1992 & 1994). Fn (z) an Fn(z) a0 bn Fn-1(z) b0 cn Fn −2 (z) The respective equation given in equations (3) and equation (4) is converted into an equivalent one-dimensional polynomial, which in turn is tested utilizing the proposed necessary conditions and Marden table (Marden, 1966). The main idea used in obtaining the onedimensional equivalent polynomial is inversion of every variable and coalition of all the inverted variables into a single variable in z with |z|<1. The resulted polynomial F(z) is handled easily by the proposed procedures. The proof for using the inverse of ‘z’ is given in Appendix. Fn-2(z) . . c0 . . F0 (z) z0 F0(z) z0 2 3 . an- an- 1 2 a1 bn- a2 bn- 1 2 b1 cn- b2 cn- 1 2 c1 . . c2 . . (m1) a1 m . (m2) a2 . . an-2 b1 an-1 b0 an . . bn-1 c0 bn . . . cn a0 The first step to formulate Marden’s table is to reverse the coefficients of the given characteristic polynomial in equation (5) i.e., n Fn ( z ) = a n z + a n −1 z n −1 + a n−2 z n−2 + ... + a1 z + a 0 ………………..……………………….…….7 Using equations (5) and (7), the first two rows are formed in Marden table given in Table 1. It is observed that when ‘z’ is inverted in the equation 3.1 Proposed Necessary Conditions Let F(z) be an n-th degree one-dimensional polynomial, n-1 1 Fn −1 (z) 3.0 PROPOSED PROCEDURE n Row/Column 1 ), equation (7) is formed. Thus, z Fn ( z ) | 1 = Fn ( z ) is obtained. (5) ( z → n-2 F(z) = Fn(z) = a0z +a1z +a2z +...+an-1z+an ………………………………………………….5 with a0>0. z→ The proposed necessary conditions for testing the equation (5) are, z The elements in the third row are calculated as given below: 28 Caledonian Journal of Engineering Volume04, Number 01, January - June 2008 bn = a0 a n −1 − a n a1 Step 4: Use z1→z2→z3→…..→zk→z and get the respective one-dimensional equation of B1(z1,z2,z3,….,zk). bn −1 = a0 a n − 2 − a n a 2 ……………..8 . . Step 5: Apply the proposed necessary condition as given in equation (6) to the one dimensional equivalents. b0 = a02 − a n2 Step6: If proposed necessary conditions are satisfied, invoke Marden table. The fourth row forms the reversal of coefficients of the third row. The above computations are repeated to formulate the other rows and this completes the formulation of the entire Marden table. Based on Marden table in Table 1, for a system to be stable, the constant terms b0, c0,…,z0 of the computed Step 7: Ascertain the stability using Marden polynomials i.e., the proposed necessary conditions are applied over Marden polynomials and if satisfied for all Marden polynomials upto second degree, then the system is stable else declare the system is unstable. Fn−1 ( z ) , Fn−2 ( z ) ,…, F0 (z) respectively, should be all positive. If any of the constant term is negative then the system is unstable. Step 8: Stop the above proposed algorithm is applied for the illustrative examples. The flowchart depicting the process of algorithmic flow is as shown in Figure 1. Instead of applying the above condition, the Marden polynomials Fn−i ( z ) where i=1,2,3,… are individually tested with the proposed conditions given in equation (6). The discussed proposed procedure is given in an algorithmic form in the forthcoming section. 4.0 PROPOSED ALGORITHM The various steps involved in the proposed algorithm are as follows: Step 1: Read the given multi-dimensional polynomial B(z1,z2,z3,….,zq) Step 2: Formulate the necessary conditions and sufficient condition for B(z1,z2,z3,….,zq) equivalent to that in the equations (3) and equation (4) Step 3: With the inversion principle, Let z1 → 1 1 1 , z2 → ,..., z q → z1 z2 zq Get all the respective inverted polynomials for each necessary condition and sufficient condition as B1(z1,z2,z3,….,zk). Figure.1: Flowchart for the proposed algorithm 29 Caledonian Journal of Engineering Volume04, Number 01, January - June 2008 5.0 ILLUSTRATIONS At z= -1, F3(-1) = 4 > 0 (even polynomial)………………………….15 The proposed algorithmic procedure is applied to the following illustrations and for |a0|<|an|; |0|<|4| The discrete polynomial with three variables from (Strintzis, 1977) is written as: Step 6: The proposed necessary condition is satisfied for F3(z) as shown in equation (15), hence Marden table is invoked as given in Table 2. B( z1 , z 2 , z 3 ) = z 3 + z1 + z12 z 2 − z 2 + 4 …..9 Table 2: Marden table for F3(z) in equation (14) 5.1 Illustration 1 Applying the proposed algorithm discussed in section 4 to equation (9), 1 0 1 4 F34(z) 4 4 1 0 0 4 1 16 0 16 -16 4 64 0 240 4 240 64 -16 F33(z) F32 ( z ) F32(z) ( z 2 = 0 and z 3 = 0) B( z1 , z 2 ,0) = z1 + z12 z 2 − z 2 + 4 0 F33 ( z ) Step 1: Read the three dimensional polynomial given in equation (9). Step 2: Formulate the necessary and sufficient conditions of equation (9), B( z1 ,0,0) = z1 + 4 F34 ( z ) ( z 3 = 0) Step 7: The second degree Marden polynomial from Table-2 is, B( z1 , z 2 , z 3 ) = z 3 + z1 + z z − z 2 + 4 2 1 2 . ……………………………………………….10 F32(z)=240z2+64z-16……..…………16 Step 3: Applying the inversion principle and formulating inverted polynomials of equation (10), i.e, Applying the proposed necessary condition to equation (16), 1 1 1 , z2 → , z3 → z1 z2 z3 B ( z1 ,0,0) = 4 z1 + 1 F32(1) = 288 > 0 z1 → F32(-1) = 160> 0 (even polynomial)………………………….17 B ( z1 , z 2 ,0) = z1 z 2 + 1 − z12 + 4 z12 z 2 and |a0|<|an| ⇒ |16| < |240| B ( z1 , z 2 , z 3 ) = z12 z 2 + z1 z 2 z 3 + z 3 − z12 z 3 + 4 z 3 z12 z 2 ………………………………………………...11 From equation (17) it can be observed that F32(z) satisfies the proposed necessary conditions declaring the given original polynomial B(z1,z2,z3) is stable. Step 4: Using z1→z2→z3→z for equation (11), i) F1(z)=B1(z1,0,0)= 4z+1……………12 Also, the third degree Marden polynomial F33(z) satisfies the proposed necessary conditions which further declares the given polynomial is stable in nature. ii) F2(z)=B1(z1,z2,0)= 4z3+1…………13 iii) F3(z)=B1(z1,z2,z3)=4z4+z3+z..........14 Step 5: Applying the proposed necessary conditions to equation (14), Step 8: Stop This conclusion is same as that given in (Strintzis, 1977). At z=1, F3(1) = 6 > 0 30 Caledonian Journal of Engineering Volume04, Number 01, January - June 2008 The proposed necessary condition is applied to equation (24): 5.2 Illustration 2 The discrete polynomial with three variables from (Strintzis, 1977) is considered here: B( z1 , z 2 , z 3 ) = z 3 + z 1 + 6 z 12 z 2 i. F2(1) = 6 > 0 − z2 + 2 ii. F2(-1) = -4 < 0………….…………………..25 …………………………………………...……18 iii. |1| < |5| Applying the proposed algorithm [Step 1 – Step 8 in Illustration 1] to equation (18), the following are obtained: The necessary conditions are satisfied for F2(z) in equation (24). B1 ( z1 ,0 ,0 ) = 2 z 1 + 1 = B1(z1,z2,z3,0) = 3. F3(z) 5z2+1………………………………………….26 B1 ( z1 , z 2 ,0 ) = z 1 z 2 + 6 − z 12 + 2 z 12 z 2 ……………………………………....………..19 With z1→z2→z, equation (19) becomes, i) The conclusion for equation (26) is same as that of the equation (25). = B1(z1,0,0) = 2z+1 F1(z) |z|<1…………………………….……20 4. F4(z)=B1(z1,z2,z3,z4)= z1 z2+z2 z3+z1z2 z3+z1 z3 z4+5z 1z2 z3z4 ………………………………………...………27 With z1→z2→z3→z4→z for equation (27), ii) F2(z) = B1(z1,z2,0) = 2z3+6 = z3+3 ………………………………………………...21 F4(z) = 5z4+2z3+2z2……...................................28 Applying the proposed necessary condition to equation (21), F4(z) in equation (28) is handled with the proposed condition, which gives, At z=1, F2(1) = 4 > 0 i) At z= -1, F2(-1) = -1+3 = 2 > 0 (odd polynomial) At z=1, F4(1) = 9 > 0 ii) At Thus for F2(z), the second necessary condition is violated, indicating the given discrete polynomial is unstable. z= -1, F4(-1) = 5 > 0 (even polynomial)………29 iii) |a0|<|an| ⇒ |2| < |5| This conclusion is same as that available in (Strintzis, 1977). iv) From equation (28), F5(z) = z2F4(z)………………………………...30 5.3 Illustration 3 F4(z) satisfies proposed necessary conditions implying F5(z) also satisfy necessary condition because F5(z) is derived from F4(z), thus it is not necessary to formulate Marden table for F4(z). From (Bose, 1974), the discrete polynomial with four variables is written as, B( z1 , z 2 , z 3 , z 4 ) = z 3 z 4 + z1 z 4 + z 4 + z 2 + 5 ……………………………….………….……22 Applying the proposed algorithm, ∀ |z| ≥ 1 1. F1(z) = B1(z1,0,0) = 5 ≠ 0 ……………………………….………...…….23 Since the proposed necessary conditions are satisfied for all necessary (F1(z) to F3(z)) and sufficient condition (F4(z) and F5(z)), the given discrete polynomial in equation (22) is found to be stable. This conclusion obtained using the proposed approach is in agreement with (Bose, 1974). 2. F2(z)=B1(z1,z2,0,0)=5z2+1………………….24 31 Caledonian Journal of Engineering Volume04, Number 01, January - June 2008 applied over it, to analyze the stability conditions. The proposed approach is simple and direct in application and illustrated with suitable examples. 6.0 DISCUSSION The salient points noted in the illustration are brought out in this section. Illustration 1 employed a three dimensional polynomial and at first, the necessary and sufficient conditions are evolved from the three dimensional polynomial. For all the necessary and sufficient conditions derived, inverted polynomials are formulated and each inverted polynomial is converted to its equivalent one-dimensional polynomial. The proposed necessary conditions were applied over these one-dimensional polynomials and if satisfied, formulated Marden table else declares the system to be unstable. From the Marden table formulated, Marden polynomials were computed, over which the proposed necessary conditions were applied and stability was analyzed. It should be noted that the Marden’s table is to be computed only if the formulated one-dimensional equivalent degree is greater than 2. For illustration 1, the proposed necessary conditions were satisfied for the one-dimensional equivalent polynomials declaring the given three- dimensional characteristic polynomial is stable. In illustration 2, the proposed necessary conditions are violated for the second onedimensional polynomial of given original polynomial, declaring the given system is unstable. Illustration 3 shows a four dimensional polynomial, wherein all of its necessary and sufficient conditions derived satisfied the proposed conditions, and declared the given system is stable. 8.0 ACKNOWLEDGEMENT The authors acknowledge with gratitude, the support and facilities provided by PSG College of Technology , Coimbatore, India and fellow research scholars to carry out this research work. REFERENCES [1] B.D.O. Anderson and E.I. Jury (1974) Stability of Multidimensional Digital Filters, IEEE Transactions on Circuits and Systems, 21(2), p. 300-304. [2] E. Walach and E. Zaheb (1982) Generalized zero sets of multi parameter polynomials and feedback stabilization, IEEE Transactions on Circuits and Systems, 29(1), p.15-23. [3] E. Walach and E. Zaheb (1982) Ndimensional stability margins computation and a variable transformation, IEEE Transactions on Acoustics, Speech and Signal Processing, 30(6), p.887–893. [4] E. Walach and E. Zaheb (1980) Sign test for multivariable real polynomials, IEEE Transactions on Circuits and Systems, 27(7), p.619–625. The main advantage of the proposed procedure is, it is simple and direct in application for any kind of multi dimensional discrete polynomial under consideration. The proposed procedure can be extended to design of unknown parameters that exist in multidimensional characteristic polynomials with minimal computational effort. [5] E. Zaheb (1984) Another simplification in Multidimensional Stability tests, IEEE Transactions on Acoustics, Speech and Signal Processing, 32(2), p.453–455. [6] L. Leclerc and P. Bauer (1994) New Criteria for Asymptotic stability of one and multi dimensional state space digital filters in fixed point arithmetic, IEEE Transactions on Signal Processing, 42(1), p.46–53. 7.0 CONCLUSION In this paper, an algebraic procedure is proposed to analyze the stability of the given multidimensional discrete polynomial. The multidimensional polynomial is converted to its equivalent one dimensional polynomial and necessary conditions and Marden table are [7] M.G. Strintzis (1977) Test for stability of Multidimensional filters, IEEE Transactions on Circuits and Systems, 24(8), p.432–437. 32 Caledonian Journal of Engineering Volume04, Number 01, January - June 2008 [8] M. Marden (1996) The Geometry of Zeros of a Polynomial in a complex variable, (NY, USA: American Mathematical Society, 2nd Edition,). ⎛ 1 ⎜⎜ ⎝ zk ⎞ ⎞ ⎛ 1 ⎟⎟ ⎟⎟ = ⎜⎜ x jy + k ⎠ ⎠ ⎝ k (a.2) Multiplying equation (55) with its complex conjugate, [9] M.N.S. Swamy, L.M. Roytman and E.I. Plotkin (1985) Planar Least Squares Inverse Polynomial and Prcatical BIBO stabilization of N-dimensional linear shift invariant filters, IEEE Transactions on Circuits and Systems, 32(12), p.1255–1259. ⎛ 1 ⎜⎜ ⎝ zk [10] N.K. Bose and P.S. Kamat (1974) Algorithm for stability test of Multidimensional Filters, IEEE Transactions on Acoustics, Speech and Signal Processing, 22(5), p.307–314. [11] N.K. Bose (1979) Implementation of a new stability test for N-dimensional filters, IEEE Transactions on Acoustics, Speech and Signal Processing, 27(1), p.1-4. ⎛ 1 ⎜⎜ ⎝ zk [12] P. Bauer and E.I. Jury (1991) BIBO Stability of Multidimensional (m-D) Shiftvarying Discrete Systems, IEEE Transactions on Automatic Control, 36(9), p.1057–1061. ⎞ ⎛ 1 ⎟⎟ = ⎜⎜ ⎠ ⎝ x k + jy k ⎛ x − jy k = ⎜⎜ k2 2 ⎝ xk + y k ⎞ ⎟ ⎟ ⎠ ⎛ x = ⎜⎜ 2 k 2 ⎝ xk + y k ⎞ ⎟− ⎟ ⎠ ⎞ ⎟⎟ ⎠ ⎛ y j ⎜⎜ 2 k 2 ⎝ xk + y k ⎞ ⎟ ⎟ ⎠ ⎞ ⎟⎟ = Ak − jBk ⎠ ⎛ x A = ⎜⎜ 2 k 2 ⎝ xk + y k where, ⎛ y B = ⎜⎜ 2 k 2 ⎝ xk + y k [13] P. Bauer (1992) Finite word length effects in m-D digital filters with singularities on the stability boundary, IEEE Transactions on Signal Processing, 40(4), p.894–900. ⎞⎛ x k − jy k ⎟⎟⎜⎜ ⎠⎝ x k − jy k ⎞ ⎟ and ⎟ ⎠ ⎞ ⎟. ⎟ ⎠ Based on the values of ‘xk’ and ‘yk’, the following cases are analyzed, Case (i) Let | xk |<1 and | yk|>1 Then, | xk2|<1 and | yk2|>>1 (a.3) Using equation (56), it can be noted that, xk2+ yk2>>1 (a.4) Also, Ak<1 and Bk<1. Case (ii) Let | xk|>1 and | yk|<1 Then, | xk2|>>1 and | yk2|<1 (a.5) [14] S.G. Tzafestas (1986) Multidimensional Systems: Techniques and Applications (NY: Marcel Dekker). [15] S.G.Tzafestas, N. Theodorou & A. Kanellakis (1988) Stability of Multidimensional Systems: Overview and New results, Proceedings of International Symposium on Circuits and Systems, p. 337-344. APPENDIX For conditions in equation (5), equation (4) is satisfied and this also indicates that, Ak<1 and Bk<1 Case (iii) Let, | xk|>1 and | yk|>1 Then, definitely xk2>>1 and yk2>>1 (a.6) Thus, Ak<1 and Bk<1 Proof for using the inverse of zk The discrete variable ‘zk’ is represented as, zk = xk+jyk for k=1 to n (a.1) and |zk| > 1. Taking reciprocal of ‘zk’ to get, 33 Caledonian Journal of Engineering Volume04, Number 01, January - June 2008 Case (iv) Let, | xk|=1 and | yk|=1 Then, xk2+ yk2=2 (a.7) which implies that, |Ak|=0.5 and |Bk|=0.5 Thus from all the above cases, it is substantiated that, if the roots of ‘zk’ lies outside the unit circle ⎛ 1 ⎝ zk then its inverse ⎜⎜ ⎞ ⎟⎟ will have roots within ⎠ unit circle. 34 Caledonian Journal of Engineering Volume04, Number 12, January- June 2008 REVIEW ARTICLE Caledonian Journal of Engineering Volume04, Number 01, January - June 2008 BEARING FAULTS DETECTION USING VIBRATION ANALYSIS AND INFRARED THERMOGRPHY TECHNIQUES Ali Mohammed Al-Khanbashi1; Khalid F. Al-Raheem2 Caledonian College of Engineering, Sultanate of Oman 1 Undergraduate Mechatronics program student, CCE, alkhanbashi2004@hotmail.com 2 Senior lecturer, Mechanical and Industrial Engineering, CCE ABSTRACT Bearings and their vibration play an important role in the performance of all mechanical systems. In many cases, the accuracy of the instruments and devices used to monitor and control the mechanical system is highly dependent on the dynamic performance of the bearings. In addition, many problems arising in machines and motors operation are linked to bearing faults. Thus, fault detection of a system is inseparably related to the diagnosis of the bearing assembly. Therefore, in this project the spherical roller bearing has been selected to study and diagnose its defects and faults using two types of condition based maintenance techniques, vibration analysis and infrared thermography. Then, these two techniques have been compared to find out which is the best one to monitor the bearing faults. KEY WORDS: Vibration analysis, Bearing fault detection, Infrared Thermography, and Short Time Fourier Transform (STFT) An international paper for (Atul Andhare and Dhanesh Manik, 2007) presented results of experiments performed towards diagnosis of defects in tapered roller bearings using vibration monitoring. The usage of time-domain expressed the roles of vibration parameters such as RMS level, kurtosis, skewness and peak to valley. They extracted results for different tapered roller bearings were as the following: the RMS levels were showing in case of bearings with roller defects, the axial vibration levels were found to be higher than corresponding radial vibration. The final result showed that the overall vibration RMS level is a better indicator of bearing defect detection than the peak value, particularly for outer race defects. 1. INTRODUCTION Maintenance includes all operations such as monitoring, inspections, adjustments repairs and /or doing whatever is necessary to keep a machine, facility, a piece of equipment or transportation vehicle in proper working order. Maintenance aims to minimize the downtime and downtime costs in order to achieve the maximum use of resources without any interruption to the production schedule. There are several types of maintenance some of them condition based maintenance. Condition Based Maintenance (CBM) is an automatic process that determines when a fault has occurred (or is going to occur) in a system, and subsequently diagnoses the cause of the fault. Choi, H. and Williams, 1989 has shown the time-domain for analyzing the vibration is more effective to extract the bearing fault feature. In this paper, Time-domain using the vibration parameters of RMS and kurtosis. With the additionally obtained the histogram is the used technique. Consequently, the purpose is to diagnose whether the bearing is faulty or healthy and that will be figured out by studying the effectiveness of RMS and kurtosis on the In this project the vibration analysis and infrared thermography techniques have been used to detect the spherical roller bearing faults located in clinker cooler fan used by Oman Cement Company. The vibration signals have been analyzed using STFT as the analyzing tool in Matlab Software. 35 Caledonian Journal of Engineering Volume04, Number 01, January - June 2008 extracted signal. The fault detection passes by some steps, beginning with getting the fault signal in time domain and ending with finding values of RMS and kurtosis using the equations and that can be done via MATLAB. PC for STFT analysis using a MATLAB code. The characteristic frequencies of the bearing have been calculated using the bearing characteristic frequencies equations (See the Appendix). The BCF for the tested bearing are: 9.12, 11.8, 7.45 and 0.434 times the shaft rotational speed in (Hz), for outer fault, inner fault, rolling element and cage fault respectively. For the applied rotational frequency of 1480 rpm (i.e. 24.66 Hz), the calculated fault frequencies are 224.8 Hz, 293.06 Hz, 183.82 Hz and 10.7 Hz for outer, inner, rolling element and cage fault respectively. The approach is applicable to diagnosing machine faults under complicated conditions such as low signal-to-noise ratio (SNR) and varying speeds. Although in this paper timefrequency analysis is carried out based on the STFT, it is believed that the proposed rationale will still hold when other sophisticated timefrequency analysis techniques, such as WignerVille distribution, Wigner higher order distribution, or continuous wavelet transform are used ( L. Zhu 2007) [2]. The bearing thermal images have been collected and recorded using infrared camera IR50 for healthy and faulty bearings in two directions: horizontal and vertical. The distance between the infrared camera and bearing location is 0.5 meters. The thermal imaging technique is an efficient tool for locating and analysing the subsurface defects in the GRP pipe. The testing results show that infrared thermography test is a reliable nondestructive method for detecting any cracks present in the GRP pipe (A.Alnoobi, 2006). The defect of spherical roller bearing because of the freedom of motion in three dimensions, analysis of the spherical roller bearing requires a total of 18 displacement and velocity coefficients (Craighead, 1992). Figure.1: Experimental Setup The purposes of this project are to diagnose the bearings faults using vibration analysis and infrared thermography techniques, and compare the data obtained to know the proper technique to be used in fault detection. 3. VIBRATION AND THERMAL IMAGES ANALYSIS RESULT The vibration results in form of STFT map for both healthy and faulty bearings are shown in Figure 2. The colour bar shows the value of vibration magnitude ranging from the maximum red color to the minimum blue colour. Red colour illustrates the low vibration magnitude and blue colour illustrates the high vibration magnitude. 2. EXPERIMENTAL SETUP The experimental setup in this project is shown in Figure 1. The clinker cooler fan supported by two spherical roller bearings with serial number of 22216E, pitch diameter of 110 mm, roller diameter of 14.5 mm and contact angle of zero. An accelerometer type CMSS2200 with magnet base attached to the bearing housing and connected with vibration analyzer type MICROLOG CMVA65 is used to measure and record the time domain vibration signal in three directions (vertical, horizontal and axial), using different bearing conditions (healthy and faulty). The recorded signals have been transmitted to The range of vibration magnitude for STFT of healthy bearing is 65 to 85 mm.sec-2, but in case of faulty bearing the vibration magnitude range is 100 to 120 mm.sec-2. The vibration magnitude for STFT map in terms of faulty bearing is more compared with healthy bearing as range of magnitude. By the examining of STFT maps the bearing fault can be detected. 36 Caledonian Journal of Engineering Volume04, Number 01, January - June 2008 The thermal images for healthy bearing are shown in Figure 3 (a, b) which show that the temperature of the bearing in both directions is 32о C. The thermal images for faulty bearing are shown in Figure 3 (c, d) which show that the temperature of the bearing has been increased to 47.8о C at vertical direction and 46.6 о C at horizontal direction. The thermal images have been captured for healthy bearing and faulty bearing using infrared camera in two directions (vertical and horizontal). The difference in temperature for each bearing depends on the distance between infrared camera and bearing location. Vertical direction (healthy bearing) Vertical direction (faulty bearing) 1 0.9 20 0.8 0 0.9 0.7 -20 0.8 0.6 -40 0.5 -60 0.4 -80 0.3 -100 20 0 0.7 -20 Frequency Frequency 1 0.6 -40 0.5 0.4 -60 0.3 0.2 -80 -120 0.2 0.1 0 -100 -140 0 0.5 1.0 1.5 Time 2.0 2.5 0.1 3 0 -120 0 0.5 1.0 Vertical Direction 1 20 0.8 0 0.7 -20 0.6 0.9 -20 -60 -60 0.3 -80 -100 0.2 -120 0.1 1.5 Time 2.0 2.5 -100 -120 0 -140 1.0 -40 0.5 0.4 0.3 0.1 0.6 -80 0.4 0.2 0 0.7 -40 0.5 20 0.8 Frequency Frequency 0.9 0.5 0 0.5 1.0 1.5 Time 3 Horizontal Direction 2.0 2.5 Horizontal Direction Axial direction (faulty bearing) Axial direction (healthy bearing) 1 1 0.9 20 0.9 20 0.8 0 0.8 0 0.7 -20 0.6 -40 0.5 -60 0.4 0.7 Frequency Frequency 2.5 Horizontal direction (faulty bearing) Horizontal direction (healthy bearing) 0 2.0 Vertical Direction 1 0 1.5 Time -80 0.3 -100 0.2 -20 0.6 -40 0.5 -60 0.4 0.3 -80 0.2 -100 -120 0.1 -140 0 -120 0.1 0 0 0.5 1.0 1.5 Time 2.0 2.5 3.0 0 0.5 1.0 1.5 Time 2.0 2.5 Axial Direction Axial Direction Figure.2: STFT maps for bearing vibration signals in different directions, Column (a), for healthy bearing, column (b), for Faulty Bearing. 37 Caledonian Journal of Engineering Volume04, Number 01, January - June 2008 *>32.0°C 32.0 32.0 30.0 28.0 26.0 *<24.0°C (b) Horizontal Direction (a) Vertical Direction *>46.6°C *>47.9°C 45.0 45.0 40.0 40.0 47.8 46.6 35.0 35.0 30.0 *<30.7°C *<28.9°C (d) Horizontal Direction (c) Vertical Direction Figure.3: bearing thermal images in vertical and horizontal directions, (a), (b) for healthy bearing, and (c), (d) for faulty bearing. 4. CONCLUSION Based on the obtained results, both the techniques can be used for bearing condition monitoring. By comparing the vibration analysis with thermal image techniques results for bearing fault detection the following points can be concluded: • The initial cost for the infrared camera is more than the vibration analysis equipments. • The application of vibration analysis technique is more appropriate for mechanical fault detection than the thermal distribution detected by the thermal camera, because the mechanical vibration movement is easier to detect using vibration analysis. • The information extracted regarding bearing condition monitoring can be more effective for processed image (two dimensions data) than the vibration analysis (one dimension) if a proper image processing technique has been used. • The analysis of one dimensional data as in the vibration signal is easier and more flexible when compared to the analysis of the two dimensional data as in the thermal images. 5. ACKNOWLEDGEMENT This is an undergraduate project conducted in collaboration with the Caledonian College of Engineering and Oman Cement Company, under the supervision of Mr. Khalid Fathi, Senior Lecturer at CCE. 38 Caledonian Journal of Engineering Volume04, Number 01, January - June 2008 REFERENCES −Outer Race Defect Frequency = n ⎛ BD ⎞ fr⎜1− cosβ ⎟ (Hz) 2 ⎝ PD ⎠ [1] A.Alnoobi. (2006), Non-destructive testing and evaluation of GRP pipe using thermal imaging technique, undergraduate project in Mechanical & Industrial Engineering, CCE Oman. n ⎛ BD ⎞ −Inner Race Dfect Frequeny = fr⎜1+ cosβ ⎟ (Hz) 2 ⎝ PD ⎠ PD ⎡⎢ ⎛ BD ⎞⎤ − Rolling Element Defect Frequency = fr ⎢1− ⎜ cosβ ⎟ ⎥⎥ (Hz) PD BD ⎣⎢ ⎝ ⎠ ⎦⎥ [2] Atul Andhare and Dhanesh Manik. (2007) Diagnosis of localized defects in tapered roller. Where PD = Pitch Diameter (mm). BD = Boll or Roller diameter (mm). Fr = rotating speed of the shaft (Hz). n = number of balls or rollers. β = contact angle (degree). 2 −Cage Defect Frequency = [3] Choi, H. and Williams, W. J. (1989). Improved Time-Frequency Representation of Multi-component Signals Using Exponential Kernels, IEEE Trans in Acoustics, Speech and Signal Processing, 37(6), p.862-871. [4] Craighead I.A. (1992), An analysis of the steady-state and dynamic characteristics of a spherical roller journal bearing with axial loading, Institute of Mechanical Engineers, Proceeding of International Conference on vibration in rotating machinery, University of Bath. [5] John S. Mitchell (1981) Machinery analysis and monitoring, 2nd Ed. (Tulsa-Oklahoma: Penn Well Publishing Company,). [6] L.Zhu, H. Ding and X. Zhu. (2007) Synchronous averaging of time-frequency distribution with application to machine condition monitoring Journal of Vibration and Acoustics, 129, p.441-447 [7] R. Barron. (1996) Engineering condition monitoring, 1st Ed. (New York: Addison Wesley Longman Inc) [8] Xavier P.V. & Patric O. Moore (2001) Nondestructive testing handbook Vol.3 Infrared &thermal testing, 3rd Ed. (American Society for Nondestructive Testing). APPENDIX Bearing Characteristic Frequencies equations: 39 fr ⎛ BD ⎞ ⎜1− cosβ ⎟(Hz) 2 ⎝ PD ⎠ ENTER TITLE HERE (14 PT TYPE SIZE, MAXIMUM OF 10 WORDS (IF POSSIBLE), UPPERCASE, BOLD, CENTERED) (Gap 12pt) Author Information entered here: Name (12 pt type font size) Affiliation Address Country (10 pt type size, upper and lower case, centered under the title) How to Use This Document Template Change the information on your document to contain the information you would like. For the body of your document, be sure to use Times New Roman 10 pt. font, upper and lower case, single spaced with a double line between the paragraphs. Allow an extra half space above a line containing superscripts and/or below a line containing subscripts. The whole text should be justified. The headings should be 12 pt size, upper/lowercase, and bold. ABSTRACT Maximum 200 words here. KEY WORDS Maximum of 6 words 1.0 INTRODUCTION Clearly explain the nature of the problem, previous work, purpose, and contribution of the paper. 2.0 BODY OF PAPER Enter your text here. Tables and Graphs: Minimum of 10 pt type size, minimum of line thickness at .013” or .30 mm, all captions should be upper and lower case, centered over 1 or 2 columns of body text. Illustrations and Photographs: Halftones, minimum of 10 pt type size, captions should be in upper and lower case, centered over 1 or 2 columns of body text. Images must be computer-designed and submitted as EMBEDDED images in your document (postscript or MS Word format). Digitized photographs in 256 grey scale are recommended. Please do not submit colour images. References should be cited as in the text based on Harvard Presentation. Example: Rogers discussed the use of LRT in five different cities (Rogers, 1978). Talvithe presented a simultaneous prediction model for LRT and bus corridors based on a logit theory (Talvithe, 1978). Zupan has developed a model based on regression analysis, to estimate future ridership of RTS in New York city (Zupan, 1978 & 1979). A disaggregated behavioural model for mode choice was formulated (Hobeika, 1974, Kocur, 1982). Chumak studied the impact of LRT on the transportation systems in Calgary (Chumak, 1984). 3.0 CONCLUSION Clearly indicate advantages, limitations and possible applications. 4.0 ACKNOWLEDGEMENT A brief acknowledgement section may be included here. Note that: A4 size in two columns ,margins left, top & bottom = 1.25 in. Maximum No. of pages should be 5-6. Paper should be submitted both in soft and hard copies. REFERENCES References must be arranged in the alphabetical order. The correct format for references is: Proceedings Papers: [1] W.J. Book.(1990) Modelling design and control of flexible manipulator arms: A tutorial review, Proc. 29th IEEE Conf. on Decision and Control, San Francisco, CA, p.500-506. Journal Papers: [2] M Ozaki, Y. Adachi, Y. Iwahori, & N. Ishii. (1998) Application of fuzzy theory to writer recognition of Chinese characters International Journal of Modelling and Simulation, 18(2), p.112-116. *Note that the journal title and volume number (but not issue number) are set in italics. Books: [3]R.E.Moore. (1966) Interval (Englewood Cliffs, NJ: Prentice-Hall). analysis Caledonian Journal of Engineering Published by Caledonian College of Engineering, CPO Seeb 111, OMAN Volume04, Number 01, January-June 2008 Papers Virtual Cells for Manufacturing Systems Under Turbulent Environment – A Review of Thrust Areas R.V.Murali…………………………………………………………………………………………...1 Removal of Copper Ions from Effluents Using Coconut Shell Coke in a Fixed Bed Adsorber S Feroz and Shah Jahan…………………………………………………………………………...7 Coast down Factor to Investigate The Tribological Behaviour of Lubricants in Journal Bearing. R. Edwin Browne, Dr. K. P. Ramachandran, Dr. A.K.M. De Silva, Prof. D.K. Harrison……..……………………………………………………………………………….…….15 Probalistic Analysis of a System with Two Types of Repairman and Patience Time Wherein the Initial Repair is undertaken by the Ordinary Repairman Vandna Bhagat………………….……………………….…….………..….………………..……21 Stability Analysis Of Multi Dimensional Discrete Polynomial Employing Equivalent One-Dimensional Polynomial Sivanandam S.N.and Rajan S……………………………………………………………...…….27 Bearing Faults Detection Using Vibration Analysis and Infrared Thermogrphy Techniques Ali Mohammed Al-Khanbashi; Khalid F. Al-Raheem .…...........................................……35