Gaussian Mixture Model and Double-Weighted Deep Neural Networks for Data Augmentation Soft Sensing

Abstract

In practice, data-driven soft sensors often face data shortages in modeling. Data augmentation technology has offered a feasible solution for this problem in recent years. However, how to better use virtual data for data augmentation is still an open topic. In this paper, a novel data augmentation soft sensing method is proposed. It uses Gaussian mixture models (GMM) to generate virtual data for the training dataset, and developed a double-weighted neural network (dwDNN) for weighted regression modeling. On top of that, the Bayesian optimization algorithm is applied to the weight selection of dwDNN to further enhance the efficiency and effectiveness of GMM -dwDNN on virtual data. In the end, a real industrial case is used to illustrate the superiority of the proposed approach in soft sensing.