Intelligent Monitoring Method For Consistency Of Plastic Injection Molding Process

The plastic injection molding process has the advantages of high precision and high degree of automation,and is the most important forming process for plastic products.The stability of the plastic injection molding process has an important impact on product accuracy and forming efficiency.However,the traditional forming quality monitoring process mainly relies on manual visual inspection and offline sampling inspection.Due to the limited quality of visual inspection,off-line sampling feedback is not timely,costly and not inefficient,and can't meet the high-efficiency and high-quality production needs of modern industry.With the widespread use of temperature and pressure sensors in plastic injection machines,the forming quality monitoring process based on the sensing information of the injection machine becomes possible.In this paper,the stability analysis method of the process data of injection molding machine is studied.The contents are as follows:Firstly,the data collected by the injection machine was analyzed,and the collection platform of injection pressure curve,screw displacement curve,screw speed curve and torque curve was established.Moreover,the influence of injection pressure,screw displacement and screw speed on product quality stability is analyzed.Aiming at the problem of large amount and high dimension of the injection machine data,this paper firstly extracts the characteristic value of 15 injection molding processes by artificial feature extraction method,and obtains the correlation between the selected and the size through correlation analysis.Through analysis,six characteristics with high correlation with size were obtained,including peak injection pressure,integral of injection section,integral of metering torque,static standard deviation of injection pressure,dynamic standard deviation of screw displacement,and dynamic standard deviation of screw speed.The fluctuation of the Euclidean distance between the characteristic values was monitored and screened to indirectly monitor the fluctuation of edge thickness.Furthermore,sparse auto-encoder and principal component analysis are proposed for dimensionality reduction.The feature values which can measure product fluctuation are extracted from the collected process data,and the weights of dimensionality reduction data in each dimension are obtained,which can be used as a basis for stability determination.The characteristic value that can measure product fluctuations are extracted from the collected process data,and the weights of the data in each dimension after dimension reduction are obtained,which can be used as the basis for stability analysis.Through the synchronous monitoring of the machine and product status,the effects of the two dimensionality reduction methods are compared.The experimental results show that the principal component analysis method is slightly better than the sparse auto-encoder.Based on the feature extraction,two methods of stability analysis of KNN and LSTM neural networks are established,and the method is implemented in Python environment.Through the comparison of the stability analysis of the forming process data and the quality inspection results of 170 samples,it is found that when the product sizes of different process groups are significant different.In the identification of unstable state,the state recognition accuracy of LSTM neural network is higher than that of KNN algorithm,but there is a certain deviation in the prediction accuracy of the samples in the group.Due to the accuracy of the sensor,the size difference caused by the fluctuation of the machine under the same process condition is difficult to distinguish.