Research And Application Of Single Crystal Silicon Scheduling Algorithm

The production plan of single crystal silicon played an important role in optimal production which contained coordinating production resources and increasing production efficiency. However, seldom single crystal silicon enterprises used intelligent algorithm to get scheduling plans. As a result, there were some defects, such as response lag, lower foresight, weaker global awareness, etc. In this paper, the key algorithm and its application were studied based on the non-steady state production environment.Different types of single crystal silicon had different output rates, which caused greater problems. In this paper, a multi-objective optimization model was proposed based on the steady state production environment. In order to predict defective rate, the BP neural network (BPNN) was used to train the historical production data. Then the improved genetic algorithm (IGA) was used to make the scheduling plan of the production resources. With abilities of parallel computing, rapid response and decreasing duplicate scheduling, the scheduling plans achieved optimal results in the completion time minimization, submission time minimization and energy consumption minimization.Single Crystal Silicon production workshop was a non-steady environment with many disturbance factors. Therefore, there were many limitations of the scheduling plans under steady state. In this paper, a hybrid scheduling model was proposed based on non-steady state. Combined with the prediction model, the proposed model applied preservation mechanism to save critical processing information. And it adopted temporary disturbance rescheduling mechanism to formulate optimal scheduling plans for completing remaining tasks. With the influence of three disturbing factors consisted of increasing temporary orders, reducing temporary orders and temporary machine fault, the hybrid scheduling plans were proved to be better than the original plans in the experiment.The knowledge summarized from the daily production information was not only benefit to the hybrid scheduling model obtaining the more accurate prediction of defective rate and more reasonable scheduling, but also benefit to managers understanding the details of workshop dynamic and changes in market supply and demand. However, the formation of this knowledge was discrete and fragmented. Therefore, a design of knowledge base system based on B/S mode and RESTful interface technology was proposed in this paper. The system not only generated scheduling plans rapidly, but also analyzed huge amounts of data to achieve the extracting, updating, and application of knowledge. And rational and orderly management of knowledge resources provided reliable decision support for the production of single crystal silicon. Through deployment and test of the knowledge base system, the results verified the stability of the system and ability of quick response and fully researched the intrinsic value of the information. Finally, it achieved the expected design goal.