Research On Queueing Network Modeling And Simulation Of The Typical Manufacturing Systems With Complex Topologies

In the planning and design stage of manufacturing systems,the optimal resource allocation problem has to meet the requirements of the system production capacity and the order due-dates.Therefore,feasible and effective methods for evaluating performance measures of stochastic manufacturing systems are prerequisite for determining whether the decision variables meet the constraints.In the context of customized equipment manufacturing industry,three kinds of manufacturing systems with complex topologies were studied and modeled as open queueing networks with finite buffers.This kind of queueing network model is difficult to analyze due to the absence of a product-form property that caused by the occurrence of blocking.Moreover,there are many limitations in the existing analytical methods.Therefore,performance evaluation approaches were proposed for each kind of system to fast calculate the system performance measures and then applied to solve the optimal resource allocation problems efficiently.This research will deepen the application of queueing network theory in the performance modeling and optimization problems in stochastic manufacturing systems.The results can provide a basis for system design and resource planning when expanding or establishing new factories or workshops in the customized manufacturing enterprises,to effectively reduce the cost of investment and achieve the expected performance.For manufacturing systems with rework processes,the system was modeled as an open queueing network of GI/G/1/K queues with blocking and feedback.An approximation method called the Rate Iterative Method(RIM)embedded with the Generalized Expansion Method(GEM)was proposed for performance analysis.Interactions between queues were described by the input and output processes and the backward processes were decomposed into splitting and merging processes and then integrated with the input and output rate at each node in the network.The RIM is a two-layer nested iteration algorithm.In the main iteration,it does the operations in an iterative way until the output rate at each queue converges.Within the secondary iteration,several nonlinear equations were established for describing the behavior of the nodes and solved iteratively for calculating the parameters needed at each node.The general algorithm was shown in detail and its convergence was verified.Finally,the simulation models were built,and the accuracy and the efficiency of the proposed method were tested by comparing the results with other analytical methods and simulation results from a series of designed experiments.Meanwhile,the factors that influence the performance of the system were analyzed and their behavioral tendencies were illustrated.For manufacturing systems with match processing constraints,the system was modeled as an open queueing network with s dual-synchronization constraint and fixed matching constraint.Three assumptions of queueing model were considered,i.e.M/M/1/K,GI/M/1/K and GI/G/l/K,and the developed Decomposition of State Space Method(DSSM'),Exponential Expansion Method(EEM')and Generalized Expansion Method(GEM')were proposed for different systems with the corresponding assumption for performance analysis.Two different types of parts are assembled and then processed on one machine simultaneously.These parts no longer have interchangeability among the same types of parts after having been processed in the matching node.They are disassembled and flowed to receive a different series of processes respectively.At last,they should be re-assembled at the final assembly node,and the two parts must be the exact pair that went through the matching node together.The matching node and final assembly node were separated into two coupled sub-nodes respectively,so the dual-synchronization constraint problem was converted into a synchronization constraint of the assembly process.Besides,the fixed matching constraint should be ensured by the assumption of first-come-first-service rule.The improved techniques to solve these problems in each analytical approach was given in detail.Finally,the simulation models were built,and the accuracy and the efficiency of the proposed method were tested by comparing the results with other analytical methods and simulation results from a series of designed experiments.Meanwhile,the factors that influence the performance of the system were analyzed and their behavioral tendencies were illustrated.For manufacturing systems with batch transfer,the system was modeled as an open queueing network with blocking and state-dependent batch transfer,and a developed Decomposition of State Space Method(DS2M)was proposed for performance analysis.Automated guided vehicles(AGV)are responsible for transporting the work-in-process between the workstations,where the transferred batch size depends on the number of jobs in the buffers and the capacity of transporters.The system was decomposed into several associated sub-systems.In each sub-system,it was modeled as a continuous-time Markov chain and state transition balance equations were generated.An iterative algorithm was developed for solving the steady state probabilities and computing the system performance measures.The state transition analysis for different types of nodes and the iterative algorithm were described in detail.Finally,the simulation models were built,and the accuracy and the efficiency of the proposed method were tested by comparing with simulations from a series of designed experiments.Meanwhile,the factors that influence the performance of the system were analyzed and their behavioral tendencies were illustrated.In the last chapter,the optimal resource allocation problem of the manufacturing system was modeled as a stochastic nonlinear integer programming model.The objective is to optimize the number of production resources so as to minimize the total investment cost,meanwhile meet the constraints of system throughput and sojourn time.The discrete polyblock method embedded with the Decomposition of State Space Method was developed for searching for an optimal solution,and its feasibility and validity of the algorithm were verified by some cases.