Research On Joint Optimization Of Production Lot And Maintenance Planning For Wafer Fabrication Considering Dependent Lead Time

It is well accepted that semiconductor wafer fabrication is one of the most sophisticated system,with the attributes of high investment of machines,complex fabrication process,multi-type of semi-finished products.In order to reduce the deprecation allowance,it is always the case that machines are in a state of high utilization level,which brings about great challenges for maintenance planners.Equipment maintenance can influence production planning in the aspect of capacity,yet traditional wafer fabrication model treat production lot size planning and maintenance planning as independent parts,which is hard to reach overall optimization.In general,lead time is regarded as an exogenous,constant factor in production planning model.However,due to the huge number of re-entrant line and stochastic factor such as machine failure,variable demand in wafer fabrication,lead time has distinguished features of nonlinear,and load dependent.Considering the nonlinear characteristics of wafer fabrication,a joint optimization model of production lot size and maintenance planning is proposed in this paper.In this paper,a function that relates expected output of a machine as a function of its expected number of failure and workload is proposed.A simulation model based on the assumption of re-entrant line,multi-type of products and time dependent failure is used to collect empirical data,which demonstrates the nonlinear relationship of the above three factors.A bivariate segmented linear regression model is proposed to obtain the piecewise linear function.The coefficient of explanation is significantly improved by the comparison with clearing function model.Subsequently,the obtained piecewise linear function of lead time is added as a constraint into the joint optimization model for multi-product re-entrant system,which is composed of two parts: production planning and maintenance planning.The goal of production planning is to minimize sum of the inventory cost,backlog cost and work-in-process cost.Preventive maintenance makes machine as good as new,while corrective maintenance makes machines as bad as old,both of which occupy the production time.It shares the planning horizon and period with production.To study the proposed planning model's performance under different system complexity level and demand mode,two models are selected for comparison,including the clearing function based separated action model and the fixed lead time based joint optimization model.Experiment results illustrate the superiority of the proposed model in wafer fabrication in aspects of both internal characteristic and external environment.