PROCEEDINGS OF THE 8TH INTERNATIONAL CONFERENCE AND 29TH ANNUAL GENERAL MEETING (YOLA 2008) THEME: AGRICULTURAL ENGINEERING AND THE CHALLENGES OF THE MILLENNIUM DEVELOPMENT GOALS VENUE: MULTI-PURPOSE HALL, FEDERAL UNIVERSITY OF TECHNOLOGY YOLA, ADAMAWA STATE, NIGERIA DATE: JANUARY 28 – 1 FEBRUARY 2008 PROC, NIAE: Volume 29, 2008 © NIAE, 2008 THE NIGERIAN INSTITUTION OF AGRICULTURAL ENGINEERS ISBN 0794-8387 FUZZY LOGIC MODELLING OF FARM TRACTOR RELIABILITY IN KWARA STATE, NIGERIA A.O. Ogunlela and T. A. Ishola Agricultural Engineering Department,University of Ilorin, Ilorin. ABSTRACT Preliminary works of data collection and analysis were carried out to examine the reliability of farm tractors. Sugeno-type Fuzzy Logic model was developed to estimate the reliabilities of farm tractors and their systems. The results show that Fuzzy Logic model is reasonably accurate in predicting reliability of farm tractors. The fuel system was observed to be the most reliable of the tractor systems. The electrical system had the greatest tendency to failure on the farm tractors surveyed. Keywords: Reliability, Fuzzy Logic model, Farm tractors. 1. INTRODUCTION As more and more capital in the form of machinery replaces manual labour on the farm, the reliability of the equipment used assumes greater importance. Indeed, deeper insight into failures and their prevention is to be gained by comparing and contrasting the reliability characteristics of systems that make up the tractor. Reliability is defined as the probability that the equipment or system will complete a specific task under specified conditions for a stated period of time (Amjad and Chaudhary, 1988). Hence, reliability is a mathematical expression of the likelihood of satisfactory operation. A failure may be referred to as any condition which prevents operation of a machine or which causes or results in a level of performance below expectation. Owing to the importance of timeliness of operations in obtaining high yields, machinery breakdown especially at busy period such as sowing or harvesting can lead to large losses of revenue apart from the cost of repairing the equipment. If estimates could be made of when equipment is likely to fail, this would assist in planning machinery purchases and spare parts inventories and reduce costs. The concept of Fuzzy Logic was conceived by Lotfi Zadeh in 1965, and he presented it as a way of processing data by allowing partial set membership rather than crisp set membership or non-membership (Kaehler, 2005).In Fuzzy Logic, an assertion is allowed to be more or less true. A true value in Fuzzy Logic is a real number in the interval [0, 1] rather than the set of two true values {0,1} of Classical Logic. Areas of application of Fuzzy Logic include cement kiln control, information retrieval systems, laboratory water level controller, etc (Jaya et al., 2004). Fuzzy Logic provides a simple way to arrive at a definite conclusion based upon vague, ambiguous, imprecise, noisy or missing input information. A Fuzzy Logic approach control problem mimics how a person would make decisions, only much faster. It uses an imprecise but very descriptive language to deal with input data more like a human operator (Kaehler, 2005). Fuzzy Logic is an important decision making tool for comparing a developed product with a similar product available in the market. This technique helps to transfer the opinion of experts as well as objective information, into numerical values which can be used for decision making (Jaya et al., 2004). Fuzzy Inference system has been successfully used in predicting many biological parameters. The use of Fuzzy Inference System is a means of modelling that allows apt description of mapping inputs to outputs (Brown-Brandl et al., 2003). Fuzzy models use qualitative relationships and can include intuitive understanding. They may be developed significantly faster than analytical models. Fuzzy models also eliminate the need of coefficients in relationships to account for uncertainty between variables (Center and Verma, 1997). Applications of Fuzzy Inference system to model complex physical and biological systems are well documented and growing (Brown-Brandl et al., 2003). The main objective of this work was to use a Fuzzy Logic model to estimate reliability of farm tractors during their operating life. This will help in assessing the suitability of imported tractors to Nigeria. Necessary steps can then be taken to make these imported tractors more reliable. 2. MATERIALS AND METHODS 2.1 Theoretical Development R(t ) 1 F (t ) Amjad and Chaudhary (1988) used Weibull failure model to apply reliability theory to farm machinery with special reference to mechanical reaper. It was claimed that the Weibull failure model is very basic and applicable to all farm machines such as combines, tractors etc. The Weibull cumulative distribution function (cdf) or failure function is given as (3) F t 1 exp [(t ) ] (1) where logarithm. exp = base of Naperian α = scale R(t ) exp [ t ] (4) For the estimation of α and β (Weibull parameters) simple regression analysis was used. This method is based on the fact that the reliability function of Weibull distribution can be transformed into a linear function of Ln t by means of double logarithmic transformation. Taking the natural logarithm of both sides of equation (2) twice, gives: parameter (months). β = shape parameter or Weibull slope (ratio). γ = location parameter or lower bound of life (months). t = time successive failure (months). between F t 1 exp [ t ] Ln t 1 Ln Ln 1 Ln 1 F (t) (6) This is of the form Y = Mx + C where (2) M = 1/ β (Slope of the linear equation) The reliability function R(t) given as is therefore (5) Amjad and Chaudhary (1988) also stated that in the case of farm machinery, the first failure can be expected as soon as the machine is placed in service. Hence, the lower bound of the life or location parameter (γ) is zero. Thus γ = 0 and the Weibull cumulative density function becomes Ln Ln 1 Ln t Ln 1 F (t) C = Ln α (Intercept of the linear equation) The Weibull model provides considerable flexibility in describing failure distribution (Amjad and Chaudhary, 1988). That is, upon setting β = 1, equation (2) becomes exponential distribution with a delay (which can be thought of as a guarantee period within which no failure can occur, or a minimum life). Thus, the assumption of a constant failure rate (exponential model) is also included as a special case in this Weibull failure model. Wingate-Hill (1981) and Amjad and Chaudhary (1988) pointed out that reliability of a machine is a product of the individual component's reliability. If the engine, transmission, tyre and steering have reliabilities of R1, R2, R3, and R4 respectively, for a specified condition, then the reliability, Rt , of the tractor as a whole is obtained as Rt = R1 R2 R3 R4 (7) 2.2 Data Collection and Analysis A questionnaire was designed to collect data on time between successive failures of each of the systems of farm tractor. The tractor systems considered were engine, transmission, hydraulic, steering, Table 1. Weibull Parameters (α and β) of Tractor Systems α β Engine 31.75 2.88 Hydraulic 25.87 4.36 Steering 19.79 2.91 Transmission 23.36 2.98 Fuel 39.31 3.39 Cooling 24.99 3.51 Traction 19.12 3.14 Electrical 18.41 3.19 Tractor System electrical, traction, cooling and fuel. A total of sixty questionnaires were administered to twenty-six organizations. The organizations were well spread all over Kwara State, Nigeria. Forty-five tractors were surveyed and they were all still serviceable. They had covered a period of operation ranging from thirty to one hundred and thirty-six months. The tractors were generally the medium power rating tractors and were used for tillage, haulage and stationary barn yard operations. The data obtained from time measurement were compiled and used to quantify reliability. The data were grouped into class intervals and analysed using median rank and regression equations in conjunction with the Weibull model. The time between successive failures data for each system of all tractors irrespective of the make was analysed to obtain the overall Weibull parameters, α and β (Table 1) and reliability of each system. With the use of equation (7), the overall reliability of the whole tractor was obtained from the individual reliability of the eight systems of the tractor considered. This computation was done for various times between successive failures. The individual reliability of the eight systems of the tractor constituted the input data while the overall reliability of the whole tractor obtained from them was the output data. Hence, sets of input/output data were generated. 2.3 Modelling The reliability input/output data set was randomly divided into two subsets. Set 1 consisted of 85 independent data points and was used in the model development. Set 2 consisted of the remaining 30 data points, and the model evaluation was completed using this set. In a detailed analysis, a random selection of 10 data points was evaluated to ensure accurate model predictions. Fuzzy Inference Systems The Sugeno-type Fuzzy Inference system was developed in this research. For a system with two input variables and one output variable, Sugeno-type fuzzy logic models have the form: If x is A and y is B, then z = f (x, y) (8) where x and y are system input variables and z is system output variable. A and B are antecedent membership functions and f (x, y) is a crisp function in the consequent. Usually f (x, y) is a polynomial function of the input variables x and y; but it could be any function as long as outputs produced by the Fuzzy Inference system can appropriately simulate the system being modelled (Mathworks, Inc., 2002). Clustering Clustering is used to establish membership functions during the development of Fuzzy Inference system. Usually, cluster definition is based on spatial distances between data. In a cluster space of numerical data, the distance between any two points is less than the distance between any point in the cluster and any point outside the cluster (Mathworks, Inc., 2002). For fuzzy clusters, the boundary between any two clusters is fuzzy. Membership functions are established by assigning a degree of belonging for each data point to a given cluster as the fulfillment of the membership function. In this research, subtractive clustering was used to establish initial Sugeno-type membership functions. Neural Network for Tuning Fuzzy Inference Systems A key characteristic of artificial neural networks is their ability to learn, a process by which a neural system acquires the ability to carry out certain tasks by adjusting its internal parameters according to some learning algorithms. The initial Sugeno-type Fuzzy Inference system was optimized by ANFIS (Artificial NeuroFuzzy Inference System) (Mathworks, Inc., 2002), in which the Fuzzy Inference System is implemented into the framework of a supervised feed-forward type artificial neural network. The training, or learning, algorithm of the artificial neural network is a combination of back-propagation and the least-squares methods. The synapses in ANFIS are not given values (synaptic weights). They only indicate flow direction of data. Parameters that are adjusted are in node transfer functions. In other words, the membership function parameters are adjusted, not synaptic weights. Fuzzy Logic Model This model, consisting of eight inputs (individual reliabilities of the engine, transmission, hydraulic, steering, electrical, traction, cooling and fuel systems of the tractors) and one output (overall reliability of the tractors), was implemented using MatLab(Mathworks, Inc., 2002), using a Sugeno-type modelling approach. The Fuzzy Logic Model was developed using the aforementioned ‘dataset 1’ as a training set, and ‘dataset 2’ as the testing set. The model was developed using ‘genfis2’ (which utilizes subtractive clustering) within Matlab (Mathworks, Inc., 2002), and ANFIS as the neural net training routine. 3. RESULTS AND DISCUSSION The description of the model was a set of five rules dictating the mapping of the inputs to the output. The five rules are shown graphically in Figure 1. Note that the inputs are combined as fuzzy clusters described with Gaussian membership functions. The model uses a logical ‘AND’ relationship to combine the data space into fuzzy clusters. The degree of belonging of an input vector to a particular cluster defines the contribution of the associated rules. The ultimate output is a weighted average of each of the contributing rules. Examining the rules presented in figure 1 reveals that the overall reliability of the tractor is sensitive to all the inputs (reliabilities of the tractor systems). The projections of each cluster on each of the tractor systems change dimensions at every rule which indicates that they all have some impacts on the output (overall reliability of the tractor). A plot of the Fuzzy Logic model estimated reliability and the reliability values obtained from reliability function equation for farm tractors (Figure 2) shows that the Fuzzy Logic model was able to capture the relationship between the eight inputs (individual reliabilities of the tractor systems) and the output (overall reliability of the tractor) with a good fit. The coefficient of determination (R2) was 0.85, the slope was 0.96 and the standard error of the estimate was 0.15. Fuzzy Logic model of each tractor system was also developed using time between failure as input and reliability of the system as output. The values of Fuzzy Logic model estimated reliabilities and the actual (calculated) values of reliability were very close. The coefficients of determination (R2) were virtually unity (0.9999). However, in order to compare the relative reliabilities of the tractor systems, the Fuzzy Logic model estimated reliabilities were plotted against time between failures as shown in Figure 3. With reference to Figure 3, the fuel system was found to be the most reliable. It has the least slope. The second most reliable system REFERENCES Amjad, S. I. and Chaudhary, A. P. 1988. Field reliability of farm machinery. Journal of Agricultural Mechanization in Asia, Africa and Latin America. 10(1):73 - 78. Brown-Brandl, T. M., Jones, D. D. and Woldt, W. E. 2003. Evaluating Modelling Techniques for Livestock Heat Stress was the engine. This is followed by the hydraulic, cooling, transmission, steering, traction and the electrical systems in order of decreasing reliability. The highest reliability value of the fuel system could be explained by the fact that routine maintenance of changing the fuel filters on the farm tractors reduces the frequency of breakdown of the system. The low reliability value of the traction system is understandable from the fact that the farm tractors in the area of study operate on rough terrains where tree stumps and other damaging materials are not properly removed. Likewise, the relatively low reliability value of the steering system is due to the fact that the tractors surveyed are made to work on rocky, difficult and rough fields. 4. CONCLUSION This paper reported an investigation to assess the reliability functions from the breakdown records of farm tractors. Sugeno-type Fuzzy Logic model was developed to estimate the reliabilities of the various systems of the farm tractor and the overall reliability of the tractors. The results show that the Fuzzy Logic model estimated reliability is reasonably accurate and is comparable to the reliability values obtained from reliability functions. Comparisons of the Fuzzy Logic model estimated reliabilities of the tractor systems showed that steering, traction, and electrical systems have higher tendency to failure than other systems of the tractor. The fuel system was observed to be the most reliable of the tractor systems. Prediction. Presentation at the 2003 ASAE Annual International Meeting, Las Vegas, Nevada, U. S. A. pp 1 - 15. Center, B. and Verma, B. P. 1997. A Fuzzy Photosynthesis Model for Tomato. Transactions of theAmerican Society of Agricultural Engineers. 40 (3): 815-821. Jaya, S., Naidu, B. K. and Das, H. 2004. Fuzzy Logic-A multi attribute Decision making Approach in Product Development and Sensory Evaluation. Presentation at the 2004 ASAE/CSAE Annual Internation- al Meeting Ottawa, Ontario, Canada. pp 1 - 8. Kaehler, S. D. 2005. Fuzzy Logics – An Introduction. http://www.seatlerobotics.org. (accessed 2006). MathWorks, Inc., 2002. Fuzzy Logic Toolbox User’s Guide Version 2. The MathWork, Inc.,2002. Wingate-Hill, R. 1981. The application of reliability to farm machinery. Journal of Agricultural Engineering. Winter edition. pp 109 - 111. Figure 1. MatLab (Mathworks, Inc., 2002) interactive interface describing “Fuzzy Logic Model.” Each row of membership functions represents a rule and consists of eight membership functions, corresponding to each input. The first eight columns represent the eight inputs, (reliabilities of the tractor systems). The last column represents the weighted output (overall reliability of the tractor) of each of the five rules.