Study On Simulation Optimization And Intelligent Prediction For Hybrid Production Of Multi-Type Products

The prediction of production time helps enterprises to grasp the situation of production,respond to the fluctuation of production performance in time,and reduce the loss caused by the fluctuation of performance of production system.According to the predicted results,production parameters can be adjusted in time,production performance can be improved and the production activities can be arranged reasonably.However,hybrid production of multi-type products is complex and highly dynamic due to the variety of product quantity,types and processing route,which makes it difficult to study on the production performance with traditional method.In order to make up for the shortcomings of existing research,this paper adopts discrete event simulation combined with machine learning method to optimize the production performance of hybrid production system of multi-products and predict the production time of products.Discrete event simulation method is used to build the simulation model of hybrid production line,and machine learning method is used to build the prediction model based on the production data obtained from the simulation of production,so as to build the prediction model for the production time of products.The work done is as followed:Using discrete event simulation method and taking a real assorted candy box production line as prototype,a simulation platform is built with the simulation software for smart factory,and the control script is designed and written to realize the simulation production,data acquisition and priority control of work in process(WIP).According to the characteristics of the simulation platform,four production control strategies,namely greedy algorithm,WIP quantity controlled algorithm,WIP time fuse algorithm and fusion algorithm,are proposed.Five sets of simulation experiments are designed to study the two key parameters of the simulation platform: allowed maximum number of WIP and the timeout fusing time .The influence of the two key parameters on the performance of the simulation production system is obtained by analyzing the simulation production data.Appropriate production parameters are chosen based on the result of simulation experiment,and simulation production data acquisition is carried out under the configuration of the parameters.According to the characteristics of hybrid production of multi-products,traditional machine learning model,multi-layer perceptron,recurrent neural networks,long short-term memory neural networks and gated recurrent unit neural networks production time prediction model are built with the collected production data.Comparing the above models,gated recurrent unit neural networks is chosen as the best prediction model.The evaluation results shows that the prediction accuracy and calculation performance of the model meet the precision and real-time requirement.