Abstract: Automobile and logistics are the two critical restructuring and invigorating industries in China. In the context of assembly plant capacity expansion, the optimization design of Third-Party Logistics Milk-Run Cross-Docking (TPL-MRCD) with the average process time of parts as performance index was researched deeply, and the simulation model which satisfied the actual logistics operation demand was established. Owing to existing of many simulation input factors, the key factors were screened by Sequential Bifurcation, and the research procedures of Response Surface Methodology (RSM) and Kriging meta-model was built for key factors. Thus the difficult optimization problem caused by multi-dimensional factors was solved.The simulation experiment calculated by investigation data showed that the logistics performance could reach an ideal level through optimal allocation of key factors; the prediction result of Kriging meta-model was better than optimum performance value of response surface. The proposed model and method could provide reliable solution for formulating TPL improve plan after expanding capacity and guide to solve the optimization issue of complex simulated system.

Key words: logistics, cross-docking, simulation, sequential bifurcation, response surface meta-model, Kriging meta-model

摘要： 为分析主机厂产能扩大对汽车零部件物流产生的影响以及第三方物流企业如何识别与优化影响其系统绩效的关键因子以满足新的物流需求，在众多汽车制造商产能扩大的背景下，以零部件平均流程时间为绩效指标，深入研究了基于第三方物流零部件循环取货越库配送物流模式的优化设计问题，建立符合实际物流运作的仿真模型。由于存在较多的仿真输入(因子)，提出首先应用序贯分支筛选关键因子，再针对关键因子建立响应面与Kriging元模型优化的研究步骤，解决了由"高维因子"导致的较难优化的问题。运用调研数据进行仿真实验计算，结果表明关键因子的最优配置可使物流绩效达到较为理想的水平；Kriging元模型的预测结果略优于响应面的最佳绩效值，该模型与方法为产能扩大后第三方物流改进方案的制定提供了可靠的解决思路，具有良好的应用价值。

关键词: 物流, 越库, 仿真, 序贯分支, 响应面元模型, Kriging元模型