Multi-Sale Modeling And Optimal Control Of Multiple Re-Entrant Complex Manufacturing Systems

Semiconductor manufacturing system is a typical representative of the multiple re-entrant complex manufacturing(MRCM) systems. With the increasingly fierce competition in semiconductor market, customer demand for personalized is becoming increasingly obvious, and the requirements of the delivery are also getting higher and higher. Customer requirements for enterprises is not only to provide the needed products, but also to meet the individualized demand of customer in the shortest time. In order to real-time response to the increasingly diversification and individuation of demand,especially for the impact of the products surging or silicon wafer increasing in a short time, also including the optimal control problem of the production system in a long time, it is very difficult to get the good optimization results using conventional technique, so, a new approach is needed to improve the performance of the production system. This paper presents a new scheduling control strategy, which is well implemented in the practical production, and provide theoretical foundation and practical value to solve the real-time dynamic control problem. The main research work of this paper includes the following aspects:1. Continuous Modeling of the MRCM system. In order to evaluate the whole performance of the MRCM system, and lay the foundation for the optimal control of the multi-product MRCM system in the long time scale, so, the establishment of continuous model should be preferred. In order to reflect the re-entrance degree of the lots influenced on the performance of the system, the re-entrant factor is introduced to update the basic continuous model in this paper.Since there are varieties of products processed in the same line during the actual production process, in order to evaluate the overall performance indicators andoptimal control of the MRCM system, in here, according to the urgency of the order, the lots in the system are divided into different processing priority, the multiple product continuous model of MRCM system is established.2. The optimal releasing control strategy of the multiple product MRCM system. For the multiple product MRCM system, the releasing strategy is used to control the system in the long time-scale. For example, meeting the seasonal demand patterns or the increase/decrease of new products. The releasing strategy can influence the throughput of the overall production system. In order to meet the market demand of different products and achieve the optimal control of the multi-product MRCM system, therefore, the optimal releasing control model of the multi-product MRCM system is established. The objective of the control model is mainly to, match between the market demand rate and the productivity,the level of the WIP and the relevant problems with the overall performance of the MRCM system, and then, improve the delivery problem of the products with different urgency.3. Modeling the continuous model with Push-Pull point(PPP) policy. In order to evaluate of the overall performance of the production system from the different time-scale and better control production process through changing the priority of the products, the continuous model based on the PPP policy is established. The key element of the modelling is to introduce a control variable--the concept of the PPP. The basic principle of the scheduling strategy is, using the push scheduling strategy at the front-end of the PPP, meanwhile using the pull scheduling strategy at the back-end of the PPP. Furthermore, a numerical example is given to analyze the influence on the system’s performance under the position of the PPP change.4. The scheduling control strategy based on the PPP. When the unexpected orders reached the wafer factory, the WIP or the wafer in the system will be a sudden increase. In the case of meeting the production capacity, in order to increase the flexibility of the complex manufacturing production system and improve the real-time response capability, a new scheduling strategy model--scheduling strategy with PPP is established. The basic principle of theproposed control strategy is to, in a certain time interval, make the WIP in the backward of the PPP and the dynamic demand equal by changing the PPP.Furthermore, a numerical example is given to analyze the influence on the system’s performance under the position of the PPP change. Meanwhile, the validity of the proposed strategy is verified by comparing with the pure push scheduling strategy and the pure pull scheduling strategy.4. The linearized feedback control strategy based on the PPP. In order to real-time response the individualized demand of customer, and deal with efficiently the unexpected order reached the fabs, the feedback control strategy policy is established based on the PPP. Firstly, the control principal of the feedback control strategy policy and the specific process of solving the model are given, separately. Secondly, through the numerical example, when the PPP is changed, the overall performance of the production system is analyzed from the different time-scales. Then, the effectiveness and efficiency of the proposed strategy is verified by comparing with the pure push scheduling strategy and the pure pull scheduling strategy. Finally, the relative error between the proposed method and the pure pull scheduling strategy has been studied further when the PPP is changed under the optimal update frequency. Meanwhile, do the further analysis of the influence of the machine utilization and the demand deviation on the simulation result.Above all, in order to deal with the urgent order, a series of methods of modeling and scheduling control strategy of MRCM system from the different time scales are proposed in this paper. This study provides theoretical basis and practical value to constantly develop and design of the new product and so on,and have important directive to improve our country’s semiconductor enterprise production-management level.