Quality prediction plays an essential role in industrial process monitoring and optimization. However, the existed methods lack of the ability to extract features in multiphase and nonlinear time series data. To improve the quality prediction accuracy and simultaneously provide some suggestion of the process mechanism, this article proposes a phased dual-attention long short-term memory (PDA-LSTM) neural network, which consists of a phase segmentation module and a prediction module. The whole series data are first divided into phases in the phase segmentation module, and individual models are subsequently constructed for each phase in the prediction module to capture distinct patterns. The LSTM network with input and temporal attention mechanism is utilized to select the most relevant variables and time steps throughout the time sequence. The proposed method is verified with a numerical example and a real injection molding process. Compared with the existed methods, the PDA-LSTM method has the advantages of higher prediction accuracy and the ability to interpret the key input variable and time points of the process data.
The significant increase in water transportation has made the navigable environment of waters more and more complicated. In order to effectively alleviate traffic congestion, carry out accurate dynamic management and effective control of ship traffic flow in busy waters, and build an accurate traffic flow prediction model as the goal, this paper uses one-dimensional wavelet analysis tools to decompose traffic flow data into trends item and interference items, by comparing the results of direct use of LSTM time series prediction and the results of wavelet decomposition and time series prediction. A new combined prediction framework based on wavelet decomposition-LSTM is constructed. Through the quantitative comparative analysis of the prediction results, it is found that wavelet decomposition of the training data before the prediction of the LSTM time series then inputting the prediction separately can significantly improve the accuracy and robustness of the prediction. The training data is decomposed into a limited number of high and low frequency components, and the stability and regularity are obtained. The sub-data group with excellent performance on the above reduces the difficulty and interference of LSTM prediction to a certain extent, thereby optimizing the overall accuracy of the prediction. Experiments show that the scheme can be accurately and efficiently applied to the prediction of ship traffic flow.