Optimization Research Of Semiconductor Workshop Scheduling Simulation Based On Neural Network

Against the background of the concepts of "Industry 4.0" and "Made in China 2025",the fourth wave of the Industrial Revolution is beginning.While the improvement of manufacturing process and manufacturing intelligence has increased manufacturing productivity,it has also made the competition in the manufacturing industry increasingly fierce.For the semiconductor manufacturing industry that mostly adopts multi-variety and small-batch production methods,how to meet the customer's personalized customization requirements and carry out effective production management in many uncertain production environments is a serious problem facing it.Production scheduling is the core of modern production management and a research hotspot in production management issues.At present,a lot of methods have been proposed to solve the production scheduling problem.Among them,the method based on scheduling rules is widely used in production management based on its simple operation and high flexibility.By combining scheduling rules with other methods,there can be Develop a reasonable production plan in a targeted manner to reduce production costs and increase enterprise competitiveness.Based on the rule scheduling,this paper takes the die change time,the flow time,the delay time as the optimization goals,and establishes a simulation model to simulate the actual production process to verify the optimization effect of different rule combinations on multiple goals.At the same time,based on the disadvantage of the slow calculation speed of the simulation model,the neural network algorithm is used to predict the performance of all rule combinations,and a batch of rule combinations with better prediction performance is selected in advance for simulation calculation,which greatly speeds up the solution speed.The specific research content is as follows:(1)Based on the discrete simulation modeling function of Extend Sim software,a semiconductor production simulation model is established in combination with the actual production environment,and a number of classic scheduling rules are arranged.A rule base is established in the simulation model for the model to call.Aiming at the uncertain factors in the scheduling,the scenario manager is used to simulate the average operation for multiple times to obtain stable performance data.The simulation model is used to calculate the optimization effect of different scheduling rule combinations to compare the advantages and disadvantages of different scheduling schemes.(2)Based on the shortcomings of the slow calculation speed of the simulation model,in the case of many rule combination alternatives,the simulation model is used to solve a batch of rule combinations as sample data to train the neural network model,and the closeness of the neural network model to the simulation results is verified Traversal prediction of all rule combinations by neural network algorithm,and select a batch of scheduling rule combinations with the best prediction results to simulate and solve the optimal rule combination in the simulation model,thereby greatly speeding up the speed of solving the most rule combination.(3)Based on the research of multi-objective optimization,evaluate the comprehensive optimization effects of different schemes on multiple performance indicators by the TOPSIS method,and the entropy method is used to obtain weights to avoid the subjective impact of artificially assigning weights,which is more comprehensive and objective.Comprehensive evaluation and comparison of different schemes.