Improved DBN Based Predict The Product Quality In Injection Molding

With As the use of plastic becomes more diversified,the requirements on the quality of plastic products have been higher.Meanwhile,injection molding technology is also gradually becoming intelligent and fine.Due to the complexity and uncertainty of the injection molding process itself,exceptional products or injection failures often emerge.At the same time,during the injection,a huge amount of input and output data will be produced.Therefore,for a complex injection molding process,it is very important to mine the large amount of data by machine learning and predict product quality for the whole injection process to be efficient and economical.The influencing factors of the whole injection molding process are complicated and featured with time sequence,bringing some difficulties to the prediction of the quality of injection products.Targeted at these problems,the main research contents and results of this paper are as follows:Based on the conditional Conditional RBM and the deep belief network model,a nonlinear Conditional RBM-based DBN neural network model in theory is designed and constructed,and the Particle Swarm algorithm is used to obtain the optimal solution of this improved DBN network structure.Experiments are carried out and the model is trained by using the injection molding monitoring data of Guangzhou Borch Machinery as the input features.The inputted time sequence features are processed by Gibbs and are sampled using contrastive divergence.The quality of injection molding products is predicted.The results show that the improved model has higher prediction accuracy than the classical DBN model.This study shows that the model constructed in this paper is good in the prediction of the quality of injection molded products.The model is also feasible and practical,and has some reference value for the quality warning of the injection molding process.