Research On Resource Allocation Optimization Of The Large-scale Manufacturing System Based On The Queueing Network Model

Two methods are required to cooperate in an iterative manner for solving the resource allocation problem of discrete stochastic manufacturing systems:one is the evaluative method.For example,the finite buffer queueing network method is responsible for solving the steady-state performance values of a manufacturing system with a given resource configuration and feeding the values into the generative method.Another is the generative method,which generate a new candidate allocation according to the performance values and the algorithm rules and inputs it into the evaluative method.Due to the resource allocation problems of the manufacturing systems are NP-hard combination optimization problems,and there is a lack of closed-form formula between decision variables and performance values.Most generative methods' computation times are exponentially increased with the scale of the systems increase,and these generative methods can only solve the specific small and medium systems in a reasonable time.Therefore,this study proposed a new highly efficient generative approach,which based on the structure of the target systems and the information of the target system provided by the queueing network model.According to the characteristics and optimization objectives of the customized discrete stochastic manufacturing system,three resource allocation problems of largeunbalanced manufacturing systems with different optimization models or different topology structures were solved.Meanwhile,for the production line with batch transportation and assembly simultaneity constraint,an open finite queueing network model was modeled and the parameter allocation method for this small-scale system was addressed.This research improves the effectiveness and applicability of the modeling and solution methods of the resource allocation problems of manufacturing systems,expanded the application level of queueing network theory in the optimization of stochastic manufacturing systems.The research results will provide effective decision support for customized manufacturing industries,as well as manufacturing industries with similar production characteristics,to continuously improve resource allocation or plan a new factory.A new generative method for allocating buffer capacity,named decomposition-coordination method,is proposed for large unbalanced production lines with unsaturated supply,whose optimization objective is to minimize the work-in-process(WIP)inventory level under a minimum throughput rate constraint.Different from other generative methods which directly optimizing the systems,the proposed method decomposes the large system into several decoupled small subsystem and adds coordination variables,optimizes these subsystems independently with the corresponding coordination variables,and updates the coordination variables by using the current buffer allocation solutions of the subsystems.After several iterations between subsystems optimization and coordination variables updating,the buffer allocation result of the original production line is obtained The effectiveness and computational efficiency of the proposed approach were verified by a series of comparative experiments with the simulated annealing algorithm.Finally,in the series topology production lines,the influences of decomposition strategies and initial value settings of the DCM on the allocation results were analyzed,and the impacts of target system parameter settings on the performances of the results obtained from the DCM were also analyzed.Based on the general handling steps of the DCM,that is to define coordination variables,construct decomposition and coordination process,a solution method was proposed for solving the joint allocation problem of buffer capacity,machine number and machine selection for the large unbalanced production lines with average cycle time constraint and average throughput rate constraint.The effectiveness and computational efficiency of the proposed approach were testified by comparing the performance of the result with other generative methods.Finally,DCM was applied to analyze the influence behavior of cycle time constraint,throughput rate constraint,machine and buffer price ratio on the simultaneous optimization of buffer and machine in the series topology production lines.For the assembly production line with batch transportation,based on the state space decomposition method,an open finite queueing network model with state-dependent batch transportation and assembly simultaneity constraint was modeled.This model was used to evaluative the average throughput rate,average WIP inventory level,and average sojourn time.Then,a buffer capacity optimization model considering the cost of WIP and the cost of vehicles was presented.The Poly-block algorithm,which was embedded with the queueing network model,was used to solve the above optimization problem of the small-scale assembly production line with batch transportation.The influence of vehicle type selection on system performance was analyzed.For the complex manufacturing system with feed-forward split/merge topology,a cost optimization model under average cycle time constraint and average throughput rate constraint was established,the decision variables consisted of buffer capacities,machine numbers,and machine types.The DCM was further extended that an efficient generative method for solving the large-scale resource optimization problems of the complex topology systems was proposed.Meanwhile,developed the non-dominated genetic algorithm II(NSGA-II)and simulated annealing algorithm for solving the problem,addressed the parameter calibration analyze of these two metaheuristic algorithms andselected sub-optimal parameters settings.The performances of the proposed three optimization algorithms were analyzed and compared under different parameters and topologies of the systems.Then,DCM was used to obtain a large amount of sample data for the analysis of variance experiments.The influence of manufacturing system parameters and topologies on the system resource allocation results were statistically analyzed.Finally,the DCM and NSGA-II were used to solve a resource allocation problem of a large-scale complex manufacturing system with feed-forward split/merge topology,batch transportation,and assembly operation.