Research On The Prediction Method Of Key Process Quality In The Production Process Of Piezoelectric Ceramics

With the development of artificial intelligence and data analysis technology,quality identification and prediction technology can effectively improve the quality of the manufacturing process.As the key process of piezoelectric ceramics,the sintering process has the characteristics of multiple parameters,strong coupling,large hysteresis and nonlinearity,so it is difficult to guarantee the qualification rate and quality consistency of the sintered product.Therefore,selecting appropriate quality identification and prediction models to predict quality indicators,and adjusting process parameters based on the prediction results can effectively reduce fluctuations in sintering quality and improve product quality consistency.Based on a comprehensive analysis of the sintering process and characteristics of piezoelectric ceramics,this article summarizes the key factors affecting the sintering quality.According to the quality indicators detection lag,the sintering quality identification and prediction model is used to identify the sintering shrinkage rate,and the volume density,dielectric loss,electromechanical coupling coefficient and piezoelectric constant are predicted.Finally,the development of the quality prediction system is realized.The main contents of the paper are as follows:Firstly,the sintering process data of piezoelectric ceramic products are processed with wavelet threshold value noise reduction,and the temperature in the high temperature zone of the furnace is determined by multi-sensor fusion.The time domain characteristics of the kiln temperature sequence are extracted and analyzed,and the gray correlation method is used to analyze the sintering process variables and quality indicators.The LSTM model(LSTM-Attention)incorporating the attention mechanism is established by selecting the temperature sequence to identify the quality of the sintering shrinkage.Random forest is used to select variables for the sintered product with qualified sintering shrinkage rate,and the optimal input variable of the indicators are determined by combining the results of correlation analysis.The indicators are predicted by the multi-core extreme learning machine model(MKELM),and the improved particle swarm algorithm(IPSO)is used to optimize the parameters of the model.And compare the performance of the IPSO-MKELM model with other prediction models to obtain more accurate prediction results.For sintered products whose predicted indicators values exceed the range of the quality standard,the sintering quality can be remedied by adjusting the subsequent process parameters,thereby further improving the sintering quality and product qualification rate.Finally,according to the current situation of the traditional piezoelectric ceramics enterprise sintering workshop,the quality prediction system of piezoelectric ceramics sintering is developed.According to different quality indicators,users can use reasonable models for quality identification and prediction analysis,and check abnormal data information.At the same time,the system can provide a basis for subsequent quality analysis and process improvement.