VMD-SEAE-TL-Based Data-Driven soft sensor modeling for a complex industrial batch processes

•A stack enhanced autoencoder algorithm based on VMD is proposed in this paper. Here, VMD is implemented by decomposing and reconstructing the original data to eliminate the noise in the data. Based on the reconstructed process data, SEAE is able to better extract the deep features of the process data and retain the original data information to the maximum extent, so as to accomplish the accurate prediction of key variables.•For the prediction of key quality variables in the target domain of industrial processes, an MMD-based transfer learning algorithm is proposed. The method is integrated into VMD-SEAE for online fine-tuning of SEAE network models to avoid model retraining. Also, the proposed method retains the source domain data feature information, which enables the transferred model to accurately predict the key variables in the target domain, thus solving the domain adaptation problem as well.•Based on two actual industrial process cases, the proposed soft-sensor modeling method is used for the online prediction of quality-related variables. The experimental results show that the proposed method has accurate prediction performance.