مدیریت تولید و عملیات (Apr 2019)
A Multi-response Simulation Optimization Based Model for Operator Allocation and Job Dispatching Rule in a Cellular Manufacturing System
Abstract
The purpose of this article is to present an approach based on simulation optimization for improving the performance of cellular manufacturing systems through optimizing operator allocation and job dispatching rules in each cell. In this study, we have considered stochastic parameters, machines’ breakdown and multiple products in order to consider the problem as close as possible to real-world situation. The presented approach is composed of Taguchi design of experiments, discrete event simulation, artificial neural networks, and data envelopment analysis. First, controllable and response variables are determined based on the objective of the study and expert judgment. Then, the design of experiments is used in order to develop experimental scenarios base on controllable variables. Furthermore, simulation is used to evaluate experimental scenarios and their related response variables. Then, in order to expand the experimental results to the whole feasible solution space, artificial neural networks is used. Finally, the optimum scenario is determined using data envelopment analysis. After determining the optimum scenario, it is compared to the present condition of the case and the improvements are determined. In order to evaluate the performance of the presented approach, a lathing factory which uses a cellular manufacturing system is considered as the case study. Introduction: Due to the fact that the high volume of manufacturing systems around the world forms the cellular manufacturing system, optimization of these lines has been of great importance and so far have been studied by many researchers in this regard. Most researchers have considered the problems in simple terms and ignored many of the assumptions. They have been optimized cellular manufacturing line problems by using mathematical modellings and meta-heuristic algorithms, but it should be noted that assumptions such as the uncertainty of problem parameters, machines’ breakdown and variable demand are among the existing and dominant conditions in cellular manufacturing problems, which, by taking them, can bring the problem as close as possible to real-world conditions, and, on the other hand, research results become more practical. Because of the complicated nature of such problems, mathematical modelling will not be efficient and useful. In this situation, simulation is one of the best approaches at hand. By using simulation modelling, it is possible to consider all parameters of the problem, stochastic, which make the model much closer to reality. The purpose of this study is to present an approach for optimizing operator allocation and job dispatching rules on machines in a cellular manufacturing ambience, in order to minimize delay costs per piece and maximizing the average efficiency of machines. Since the model of this study is seeking multiple objectives, the simulation model of the problem includes several responses. In the end, the operator's optimum number for allocation to each cell and the optimal job dispatching rules in each cell will be determined with the aim of achieving the objectives of the problem. Azadeh et al. used fuzzy data envelopment analysis (FDEA) and computer simulation to optimize operator allocation in a cellular manufacturing system. They indicated the effectiveness and superiority of the method through a practical case study (Azadeh et al., 2010)(Azadeh, 2010 #9;Azadeh, 2010 #9). Besides, an approach for multi-response optimization problem by using artificial neural network (ANN) and data envelopment analysis (DEA) is studied by Bashiri et al., (2013). Studies have been done so far show that optimal operator allocation along with the optimal job dispatching rules in the cellular manufacturing system has not been performed in the stochastic conditions, and from this point of view, the present study is unique. Materials and Methods: This section describes the proposed methodology which is illustrated in Figure 1 Results and Discussion: In the present study, the cellular manufacturing system was first evaluated and the data needed to simulate the system were collected. After the initial simulation of the manufacturing system in ARENA simulation software, controllable variables were determined according to the features of the manufacturing system. Then, using Taguchi’s experimental design method in Minitab software, experimental scenarios were designed by various combinations of controllable variables. Then, the simulation model was modified and simulated according to any experimental scenario, and the problem response variables, that were the same problem objective functions, were extracted. After extracting the results of the experimental scenarios, considering that without evaluating other not tested scenarios, it is impossible to identify the optimum scenario, by using artificial neural networks, the experimental results were expanded to the entire possible modes. For this purpose, data on experimental scenarios with their results were placed as training data in the neural network. After setting the parameter, the optimal neural network was identified. Table (1) shows that the network number 7 with 6.8% error is chosen as the optimal structure of the neural network. Integrated modelling for resource allocation and job dispatching rule problem •Determine response variables and controllable variables •Gather required data for simulation •Use discrete event simulation method to model the problem •Validate the model by comparing simulation results with real data gathered from the case Experimentation phase •Determine levels of considered controllable variables •Design of scenarios by using Taguchi method •Simulate designed scenarios and obtained response variables for each scenario Estimate all possible scenarios by the combination of controllable factor levels •Train MLP artificial neural network for each response variable separately under a different structure •Determine the best ANN structure for estimating each response variable •Estimate all possible scenarios using determines best ANN structure Evaluate the efficiency of all scenarios using DEA •Calculate normalize SN ratio for estimated responses •Apply the input-output oriented model of DEA to determine efficient scenarios using estimated responses Fig. 1- A schematic view of the proposed approach Table 1- The error rate of each of the MLP ANN structures Network number Training function Number of HLs Error rate 1 BFGS 1 14.1 2 LM 2 15.4 3 GDA 2 8.8 4 BFGS 2 16.6 5 GDA 1 14.6 6 OSS 2 9.8 7 LM 2 6.8 8 LM 2 12.6 Then by using the trained network, the response variables of other untested scenarios were anticipated. The next step was to identify the optimal scenario. To do this, the Sexton oriented input-output data envelopment analysis model (Sexton et al., 1986) was used. Finally, the optimal numbers of operator allocation to each cell and the job dispatching rules in each cell were determined. Conclusion: In this study, given the existence of 5 types of the piece in the manufacturing system under consideration, minimizing the waiting time for each type of piece, plus the numbers of operators’ allocation, were the objectives of the problem. In order to determine the amount of improvement achieved, the current situation compared the manufacturing system with the optimal scenario. With estimation of the expected improvement rate, the results showed 6.972 minute reduction of waiting time for the piece of type 1, 6.818 minute reduction of waiting time for the piece of type 2, 6.03 minute reduction of waiting time for the piece of type 3, and 9.748 minute reduction of waiting time for the piece of type 4. Also, the total number of allocation operators to all cells is reduced 1 pcs, which causes a 6.25% rate reduction in the cost of human resources, however, in this case, the waiting time for the piece of type 5 increases 2.586 minute. References Azadeh, A., Anvari, M., Ziaei, B., & Sadeghi, K. (2010). An integrated fuzzy DEA–fuzzy C-means–simulation for optimization of operator allocation in cellular manufacturing systems. The International Journal of Advanced Manufacturing Technology, 46(1), 361-375. Bashiri, M., Farshbaf-Geranmayeh, A., & Mogouie, H. (2013). A neuro-data envelopment analysis approach for optimization of uncorrelated multiple response problems with smaller the better type controllable factors. Journal of Industrial Engineering International, 9(1), 1-10. Sexton, T. R., Silkman, R. H., & Hogan, A. J. (1986). Data envelopment analysis: Critique and extensions. New Directions for Program Evaluation, 1986(32), 73-105.
Keywords
