Research on Pattern Recognition Performance of Control Chart Based on Deep Learning

This paper is aiming at the problem that the existing anomaly discrimination methods of control chart can not realize the discrimination of complex anomaly data and the low level of intelligence. It is exploring the performance of control chart pattern recognition based on deep learning. It briefly introduces 1DCNN, LSTM and BiLSTM then adopts a neural network model based on 1DCNN+BiLSTM. The Monte Carlo method is used to generate the simulation data of the control chart, and different abnormal data are generated by changing the parameters for simulation experiments. According to the effect of control chart pattern recognition under different abnormal data, the training samples are optimized and the parameters of the network model are determined. The results show that the proposed method performs better in recognition.