Robust Optimization And Design For Supply Chain Based On Kriging Meta-Model

With the rapid development of economic globalization and the increasingly fierce market competition,the competition among corporations has evolved into the competition among supply chains.The system fluctuation caused by the uncertain factors existing in supply chain seriously affects the robustness of supply chain and brings a great deal of economic losses.How to develop the optimal parameter design under the uncertain environment for ensuring the normal operation and benefits of supply chain system has become one of the hot topics in the field of supply chain management.In the paper,supply chain is regarded as a complex dynamic multi-stage process from the perspective of system variation.A robust quantitative optimization methodology under the framework of Kriging meta-model is proposed to reduce the uncertainty impact on supply chain system.The proposed method can not only improve the efficiency of quality design for supply chain system,but also assure the robustness of supply chain system.Therefore,the robust parameter design of supply chain based on Kriging meta-model has the important theoretical and practical significance in the field of supply chain management.In the paper,taking the parameter design of supply chain as the subject of the research,we systematically study the robust optimization of supply chain based on robust parameter design.Specifically,we synthetically use Kriging meta-modeling,simulation experiment and empirical research,non-parametric bootstrap sampling and heuristic optimization approach.The thesis includes the following main contents:(1)Robust parameter design of supply chain under uncertainty conditions.Taking the effects of uncertainties on supply chain performance into account,two types of robust optimization approaches that combine robust parameter design with polynomial regression and Kriging metamodels are presented respectively.And the robust operation conditions for input parameter are obtained by the proposed methods for single performance supply chain.What's more,the comparison of the two optimization methods is also illustrated.The results show that Kriging method is superior to polynomial regression.For multiple performances of supply chain,the overall desirability robust optimization method based on polynomial regression and Kriging metamodel is constructed under the framework of the desirability function.And the optimal parameter level of the two optimization methods is determined.Based on the optimization results,the non-parametric bootstrapping is used to analyze the advantages and disadvantages of the two methods.The simulation results show that optimization strategy based on Kriging metmodel can effectively deal with supply chain optimization,and improve the supply chain robustness.(2)Robust design of supply chain ordering strategy with stochastic parameter.A robust optimization method which combines Taguchi's methodology with Kriging meta-modeling technique is developed for the uncertainty of order cost of Economic Order Quantity Model(EOQ)in supply chain.And this method is different from the tradinational optimization method based on game theory.Two Kriging metamodels for mean and standard deviation of total cost are constructed respectively.The robust optimum ordering strategy is obtained by minimizing the mean while satisfying a constraint on the standard deviation.In order to estimate the variability of the original Pareto frontier,a nonparametric bootstrapping prodecure is used to obtain the confidence interval of the optimal solution.(3)Robust parameter design of supply chain system considering risk-averse characterics.Aiming at the risk-averse optimization problems of supply chain system,a novel and robust approach combining Taguchi's robustness with Kriging meta-model is proposed based on the theory of the conditional value-at-risk criterion.The Kriging meta-models are used for fitting both the mean and the conditional value at risk of the multiple responses,and then the formulation for the risk-averse simulation optimization problems with multiple responses is obtained.Furthermore,Pareto frontier is obtained by combining Kriging with nonlinear programming.The bootstrapping sampling method is used to estimate the variability of the Pareto frontier as a consequence of the uncertainty in environmental variables.Meanwhile,the effect of different risk-aversion parameter on the estimated Pareto frontier is also analyzed.And the optimization results show that the proposed method is effective,sound and can provide the corresponding theoretical basis and technical support for the decision makers.(4)Robust parmeter design of supply chain system considering correlation between multiple performances.The principal component analysis method is used to transform the location and dispersion of each performance into the principal component score.In view of the double response surface method,the Kriging meta-models are constructed to describle the relationship between the decision factors and the location propertys and dispersion property value of each response respectively,then the optimization strategy based on the weighted sum of the composite score through Kriging meta-models is given to determine the optimal levels of these decision factors.At last,the effectiveness of the proposed optimization strategy is illustrated by a simulation case.The results show that the optimization strategy has fully considered the influence of the system varibility on the optimization results,and has more advantages compared to some other optimization methods.Finally,the thesis also discusses some challenging topics which deserve further research in the future based on the above research results.