FA-SconvAE-LSTM: Feature-Aligned Stacked Convolutional Autoencoder with Long Short-Term Memory Network for Soft Sensor Modeling

Ping Wu, Zengdi Miao, Ke Wang, Jinfeng Gao, Xujie Zhang, Siwei Lou, Chunjie Yang

Abstract: Highlights•A feature alignment strategy is developed to align spatial features from the stacked convolutional autoencoder, preserving temporal consistency.•A spatio-temporal model integrating a feature-aligned stacked convolutional autoencoder and LSTM is developed for soft sensors, simultaneously capturing spatial and temporal dependencies in industrial processes.•The proposed method’s effectiveness and superiority are demonstrated through comprehensive experiments on the sulfur recovery unit and multiphase flow process.