Multirate Nonlinear Process Fault Detection Based on Multiscale Hierarchical Variational Autoencoder
Bingbing Shen, Jinchuan Qian, Zeyu Yang, Le Yao
Abstract
In industrial processes, data is frequently gathered at various sampling rates, influenced by factors such as the diverse characteristics of process variables and the use of different sensors. However, the fault detection model for multi-rate data faces challenges in handling irregular data, complex modeling, and poor generalization capabilities. To effectively utilize and regularize multi-rate data and enhance detection models, this paper proposes a multi-scale hierarchical fault detection (MsHFD) model. This innovative model integrates and reconstructs irregular multi-rate data into multi-scale data, progressively advancing through different scales, and employs variational autoencoder (VAE) for nonlinear feature extraction. The hierarchical model not only preserves data change trends but also incrementally incorporates variable relationships, thereby improving detection efficiency and providing a foundation for faulty variable identification. Finally, comparative experiments with traditional linear models and non-probabilistic deep models demonstrate that the proposed model in multi-phase flow process achieves superior fault detection performance.