Process Quality Prediction Method Based on CNN_LSTM Hybrid Neural Network Model

With the continuous improvement of workshop production process, the complexity of workshop production is also increasing, which aggravates the difficulty of workshop process parameter prediction. In view of the above problems, this paper proposes a process quality prediction method based on a hybrid model of Convolutional Neural Network (CNN) and Long Short Term Memory (LSTM). The data set features composed of massive process parameter data, quality parameter data and date information were analyzed for correlation. After filtered the features most relevant to the quality indicators of the target variables, the features were used as input to construct a continuous feature map according to the time sliding window. Firstly, CNN were used to extract the feature vector, which was constructed in the form of time series and used as the input data of LSTM network, and then LSTM network was used to predict the process quality. Using the proposed method to predict the production line data of a workshop in Southwest China. The experimental results show that the material moisture and environmental temperature and humidity have a great impact, so the data set is divided according to the season to verify the model. Compared with the traditional prediction method, the Random Forest model prediction method and the standard LSTM network prediction method, the prediction method proposed in this paper has higher prediction accuracy.