Key Parameter Prediction And Scheduling Algorithm For Semiconductor Wafer Fabrication System Under Uncertain Environment

As the role of information technology and industry enhanced constantly in social development and the concept of "Internet+" growing prevalent in various industries, the semiconductor manufacturing industry is becoming more and more important. There are much uncertain production factors in a semiconductors wafer fabrication system (SWFS). With the rapid changes in market demand and competition, how to effectively improve performance index and flexibility of scheduling algorithm become serious problem. Bottleneck equipment and job cycle time are the key factors for on-time delivery rate and yield and so on, however, the identification and prediction methods of key factors are difficult to adapt for the dynamic production environment. Thus, with dynamic characteristics and bottleneck equipment, this paper focuses on parameter forecasting and scheduling problems about bottleneck equipment and job cycle time, which provides new ideas for solve system scheduling problems.(1) In view of dynamic characteristics and bottleneck drift caused by uncertain factors in a SWFS, this paper proposes sensitivity analysis based on Fourier transform method to optimize structure of the BP neural network for predict bottleneck equipment, so that the model can automatically adjust to adapt to the dynamic production environment, combined with the results of bottleneck prediction to analyze drift trend.(2) To solve the problems of mass redundancy production data and low on time delivery rate in a SWFS, the improved CLARA clustering algorithm is presented to simplify production data, while taking advantage of mutual information evaluation criteria to calculate the contribution matrix for measuring the network structure, and thus forecast local wait time and estimate the overall cycle time.(3) In view of large scale, computation complexity for SWFS under dynamic production environment, the method of machine layer decomposition based on hierarchical prediction iterative thinking is applied to subdivide scheduling problem, reduce scale of problem, update release strategies and dispatching rules based on heartbeat mechanism with hierarchical dynamic scheduling algorithm, and improve scheduling system adaptability.