Cyclic material supply system optimization for mixed model assembly lines

Developments in information technology have allowed customers to choose and get consumer products more quickly and easily from many similar products. For this reason, companies need to increase their competitiveness in order to be successful in the global market and to maintain this success. In order to increase this competitiveness, they have to improve their sales and marketing activities and should keep operating costs as low as possible. "Lean Manufacturing" philosophy, which was developed by Toyota and applied successfully in many fields in both production and service sector in recent years, also supports companies to keep these operating costs to a minimum. The main purpose of lean production is to minimize loss in the system by minimizing the activities that cause time and cost wastefulness in production or service systems. One of the sources of wastefulness in companies are logistics operations. The efforts to prevent wastagefullness arising from logistics activities revealed the concept of Lean Logistics. In this thesis, the problem of transporting the required materials from warehouse to assembly stations for mixed model assembly lines is discussed. For this problem, a material feeding system has been developed with milk-run, which is one of the lean logistics methods. This system includes the use of milk-run transport systems in order to deliver small amounts of high variety parts to required assembly line station in a timely and low cost. The studies in literature indicated that the use of milk-run trains in this area makes the material handling system simpler and provides ease of application and control in real life. In this thesis, an original mixed integer linear mathematical model has been developed for the material feeding problem in mixed model assembly lines. One of the most important features of the developed model is that it is a linear model that calculates vehicle routes and vehicle cycle times simultaneously. The considered problem is NP-Hard because it involves the Travelling Salesman Problems (TSP) which is also NP-Hard. Therefore, no solution was found for the large-scale problem with the developed mathematical model. For this reason a math-heuristic method has also been developed. Developed mathematical and math-heuristic models require, minimum and maximum number of vehicles value and minimum and maximum vehicle cycle time values as input parameters. There is no study in the literature for the calculation of these values. Within the scope of the thesis, functions that allow the calculation of the minimum and maximum vehicle numbers and the minimum and maximum vehicle cycle times theoretically were defined and contributed to the literature. The numerical analyses showed that the math-heuristic model was not sufficient for the solution of the real life sized problems. For this reason, Artificial Bee Colony (ABC) algorithm, which is a meta-heuristic approach, has been developed in order to solve the big problem. In addition, parameter analysis was performed for the developed YAK algorithm and the appropriate parameter levels were determined. The reasons why the YAK algorithm is preferred can be explained as follows; The performance of this algorithm on assembly line studies has been proven by studies in the literature. And also the algorithm can be reach effective solutions even for large-scale problems in a short time, thanks to its parallel search structure. The developed meta-heuristic method has contributed to the literature with the introduction of the ABC algorithm for multi-part, multi-vehicle mixed assembly lines with a general assignment structure that addresses the simultaneous determination of both vehicle routes and cycle times. Another issue addressed in the thesis is the development of a multi objective linear mathematical model for assembly line feeding system problems. In the developed model, the purpose functions are determined as minimization of the maximum amount of buffer stock and maximization of the lowest vehicle occupancy rate. There is no multi-objective model in the literature on this topic. Therefore, the study contributes to the literature with this aspect. In the implementation phase of the study, solutions for the assembly line material feed system of a white goods manufacturer operating in Turkey were searched. The company has a system where 15 different assembly parts are taken from the central warehouse to 28 different stations. In order to solve the system, it was tried to decide how many vehicles will be used in the system and how many minutes they will be served and which routes will be followed. The objective function for the problem obtain by: total transportation, order and hand-holding costs minus total station benefit. Because of the problem is NP-Hard, developed mathematical model could not provide a solution for a large-scale problem with 15 parts and 29 stations at an acceptable time. Because of this, the real-life problem is divided into seventeen different sub-problems. The mathematical model achieved the optimum solution for the first three sub-problems in acceptable time. And for larger size sub-problems, firstly the mat-heuristic model was tried to used. However, this method was able to produce solutions up to the ninth sub-problem and could not find solutions for larger problems either. Therefore, the ABC algorithm was used to solve the remaining problems. This meta-heuristic algorithm has been able to produce solutions for all problems in a very short time. After we obtained solutions, studies were performed to evaluate the quality of the solutions obtained by the Meta-heuristic method and the performance of the algorithm. In this context, the solutions obtained by the algorithm with different parameter values were analyzed with SPSS - ANOVA program. Then the problems are solved with the parameter values that give the best result. The results obtained for the first nine problems were compared with the results of other methods. The meta-heuristic algorithm has achieved the same results with other methods for each of these nine sub-problems. This has proven the performance of the meta-heuristic method. In the other hand, the developed multi-objective model has achieved the optimum solution only for the smallest problem. It achieved three pareto solution point that could not be superior to each other. However, it could not provide solutions for larger problems. As a result, with this study, a method that can provide quality solutions in short time for the actual material feeding system problem of a white goods manufacturer company has developed. However, it is also needed to provide multi-objective solutions for real-life problems by developing meta-heuristic methods for the multi-objective models in future studies.