Variational Autoencoder Bidirectional Long and Short-Term Memory Neural Network Soft-Sensor Model Based on Batch Training Strategy

Long and short-term memory (LSTM) has been used in soft-sensor modeling of industrial processes in recent years. However, LSTM still has many defects for soft-sensor. This article proposes a variational autoencoder bidirectional LSTM soft-sensor modeling method based on batch training (Bt-VAEBiLSTM). First, the training samples are divided into multiple batches according to the time series, in order to reduce the influence of abnormal points and noise, the variational autoencoder is then used to reconstruct the training samples in each batch in order to solve the problem of the global LSTM model discarding critical data information during training; this article proposes a batch training method that is to say the reconstructed samples are trained in batches according to the time series. After the training of a batch samples is completed, the structural parameters of the previous local bidirectional LSTM (BiLSTM) model are shared with the next local BiLSTM model as the initial parameters to retain important state information. At the same time, in order to prevent the Bt-VAEBiLSTM model from overfitting, the L2 regularization term is introduced in the loss function. The effectiveness of the proposed method is verified by simulation experiments on the grinding and classifying process.