Dynamic nonlinear batch process fault detection and identification based on two‐directional dynamic kernel slow feature analysis

Abstract

The batch process generally covers high nonlinearity and two-directional dynamics: time-wise dynamics, which correspond to inherently time-varying dynamics resulting from the slowly varying underlying driving forces within each batch duration; and batch-wise dynamics, which are associated with different operating modes among different batches. However, most existing dynamic nonlinear monitoring methods cannot extract the slowly varying underlying driving forces of the nonlinear batch process and rarely tackle the batch-wise dynamic characteristics among batch runs. In order to address these issues, a new monitoring scheme based on two-directional dynamic kernel slow feature analysis (TDKSFA) is developed by combining kernel SFA with a global modelling strategy. In the TDKSFA method, kernel SFA is integrated with the ARMAX time series model to mine the nonlinear and timewise dynamic properties within a batch run due to its capability of extracting the slowly varying underlying driving forces. Furthermore, the global modelling strategy is presented to handle the batch-wise dynamics among batches by calculating the total average kernel matrix of all training batches. After the slow features are extracted, Hotelling's T 2 and SPE statistics are built to detect faults. To solve the issue of fault variable nonlinear identification, a novel nonlinear contribution plot inspired by the pseudo-sample variable projection trajectories in the TDKSFA model is further proposed to identify fault variables. Finally, the feasibility and effectiveness of the TDKSFA-based fault diagnosis strategy is demonstrated through a numerical system and the penicillin fermentation process. K E Y W O R D S batch process, fault diagnosis, slow feature analysis, two-directional dynamics 1 | INTRODUCTION Batch processes are becoming increasingly important in the pharmaceutical, semiconductor, polymer, and biochemical industries to manufacture high value-added products. However, it is difficult to establish a batch process monitoring model due to the high nonlinearity in time-wise and the time-varying dynamics both time-wise

Developing accurate data-driven soft-sensors through integrating dynamic kernel slow feature analysis with neural networks

“…Although KSFA has been incorporated into several recent works on process monitoring (Zhang et al, 2017) and process fault detection (Zhang et al, 2021;Zhang et al, 2018;Zhang et al, 2015), the effect of the nonlinear slow features from KSFA as inputs to a nonlinear regression model has not yet been investigated. Soft-sensor design differs from process monitoring and other process applications by using regression/machine learning-type methods to create predictions of difficult to measure key product quality variables through other easy to measure process variables.…”

Comprehensive monitoring of industrial processes using multivariable characteristics evaluation and subspace decomposition

“…5. WNM x i , x j À Á of process variables in X G and X N can be computed using Equations ( 4), ( 5), and (9). NM m N and NM m G of X G and X N can be computed using Equations ( 10) and (11).…”

“…[1][2][3][4][5] Process monitoring is an effective means to detect abnormal working conditions and improve products quality. [6][7][8][9][10] The existing methods are usually divided into the model-based and data-based methods. [11,12] The model-based methods are generally difficult to be applied due to the complexity of industrial processes and incomplete prior process knowledge.…”

Slow‐varying batch process monitoring based on canonical variate analysis

“…[ 1 ] Hence, establishing an effective monitoring system for trustworthy fault detection and diagnosis is important to ensure the safety of batch production processes, which has drawn increasing attention from many researchers and practitioners. [ 2–4 ]…”