Research On Quality Prediction Method Of Die Casting Smart Factory

With the high development of production technology,manufacturing industry began to enter the industrial 4.0 era,and production data began to show the characteristics of "a large number of high-speed,diverse,low-value density" big data.The improvement of production efficiency brings new requirements for the quality prediction of manufacturing products.Similarly,manufacturing factories need to apply highly targeted quality prediction methods to meet the challenges of the times.In order to meet the needs of factory intellectualization,the production activities of the factory are combined with the relevant production data and the collection,storage and utilization of large industrial data are realized to achieve higher production efficiency and lower production cost standards.The specific contents of this paper are as follows:Aiming at the characteristics of time series,high dimensionality,heterogeneity and mass of production process data of cylinder block die casting,a quality prediction method based on LSTM neural network is proposed and compared with the traditional time series models(ARIMA and HMM).Referring to the historical data of Yingkou H Company,this paper compares the advantages and disadvantages of the above three methods for predicting the quality of cylinder block,and completes the classification prediction and regression prediction of the quality of cylinder block.It can be concluded that ARIMA model has a small amount of calculation and a fast operation speed,but its prediction accuracy is low;HMM model has a high speed and accuracy,and can be used to solve the problems of classification prediction and regression prediction,but the large increase of data volume can not significantly improve the prediction accuracy,so it can not use the value of industrial large data;LSTM model has some defects such as over-fitting,slow training speed,but pre-emption.The measurement accuracy is high and the value of large data can be fully utilized.Large data analysis platform can effectively improve the training speed of LSTM.Based on Cloudera commercial version Hadoop software,this paper builds a large data analysis platform in the laboratory environment,deploys and runs the quality prediction method of cylinder block based on LSTM.By comparing the results of calculation with that of single computer,the effectiveness,superiority and necessity of large data analysis platform are proved.The construction of cylinder block quality prediction and large data analysis platform is an important link to realize the intellectualization of cylinder block die casting plant.Based on this link,a data-driven operation mode of cylinder block die-casting plant is proposed in this paper.Through on-site investigation,the value stream of cylinder block die-casting factory is obtained,and on this basis,a complete intelligent solution of cylinder block die-casting factory is given.