Solution methods for integrated production and outbound distribution scheduling problem with multiple plants

Nowadays, companies must carry out production and distribution operations together to compete with other companies. While the production activities include the order of operations of customer demands within the facility and the scheduling of these operations, distribution activities include the process of delivering the products to customers. One of these operations is time sensitive products. Time-sensitive (perishable) products must be delivered to the customer in a limited time, after their production is completed. Newspapers, food products, ready mixed concrete mixes, nuclear medicine, and industrial adhesive materials can be given as examples of application areas of an integrated approach for perishable products. Although production planning and distribution planning are studied separately for many years, the integrated production and distribution scheduling problem has been studied extensively for about 20 years. Perishable products are not considered in most of these studies. However, the planning of the production and distribution of perishable products is more important because of it will cause the products to lose their value partially or completely in delays. In this thesis, the problem has been examined called as Integrated Production and Outbound Distribution Scheduling (IPODS) Problem in the literature. This problem includes machine scheduling and vehicle routing problems and has many variants. In this study, routing decisions are also considered allowing to visit more than one customer during the same tour. In this thesis, IPODS problems that have not been dealt with before in the literature have been examined. As a first problem, the Integrated Production and Outbound Distribution Scheduling Problem with Multiple Plants and Single Vehicle (IPODS_MP_SV) is discussed. In this problem, there are more than one plant in the system and customers are served with a single vehicle in each facility. A mathematical model is developed to solve the problem firstly. The Variable Neighborhood Search (VNS) Algorithm has been developed to obtain optimal or near-optimal solutions in a shorter time for the solution of large-scale problems. Developed mathematical model and the VNS algorithm are evaluated with 6 different parameters and compared the results. For the developed 2592 test problems, although the mathematical model could not be solved 833 test problems, the VNS algorithm found a feasible solution for all problems. Optimal results are obtained in 74 test problems with the mathematical model, and all these problems are in the case of 10 customers. The average solution times for problems with optimal results is 3085 seconds. The results obtained with the VNS algorithm, and the mathematical model were obtained in less than 1 seconds. When the results of the mathematical model and the VNS algorithm is compared, the difference between the GAP values is calculated as 15.16% on average. While the time required for the solution of the mathematical model is 7083 seconds on average, this time is about 70 seconds in the VNS algorithm. The VNS algorithm provides much better results both in terms of time and performance. Another study of the thesis is The Integrated Production and Outbound Distribution Scheduling Problem with Multiple Plants and Multiple Vehicles (IPODS_MP_MV). This problem includes many plants in the system, and more than one vehicle in these plants. For the solution to the problem, firstly a mathematical model was developed as in the first problem. The developed mathematical model is strengthened by the valid inequalities. Since the IPODS problem is known to be NP-hard class, the Memetic Algorithm (MA) has been developed to obtain optimal or near-optimal solutions for large-sized problems in a shorter time. Test instances for the IPODS_MP_MV are examined for different parameters and the results are reported. Optimal or feasible solutions are obtained in 1283 problems of 2592 test problems. The GAP value obtained according to the mathematical model was calculated as 0.48%. The GAP value is very close to "0" indicates that these results very close to the optimal result are obtained. The average time to obtain these results is 4920.55 seconds in mathematical model. Optimal solutions were obtained in an average of 538.98 seconds in mathematical model. The average time required to obtain these solutions with the MA is 3 seconds. This shows how fast and accurate results the algorithm gives. Considering that the smallest values obtained with MA, MA gives 0.08% better results. Considering that the average of 5 different solutions obtained with MA, it can be said that the mathematical model solutions 0.01% better than MA solutions. It is possible to say the same or very close results in a very short time with MA. The problems are solved by the MA algorithm at least 0.01 seconds, at most 26 seconds. In addition, while a feasible solution is obtained for all problems with MA, there are 1309 problems that cannot be solved with the mathematical model. Therefore, it can be said that the MA algorithm is more efficient than the mathematical model. As the number of customers increases, while it becomes difficult to obtain a feasible solution with the mathematical model, obtained much better results with MA. With the mathematical model, a feasible solution was obtained for 5 problems within the case of 50 customers, and the GAP value of these problems was calculated as 11.03%. 7.17% better results were obtained for these 5 problems with MA.