Improved Process Monitoring Scheme Using Multi-Scale Independent Component Analysis

Abstract The task of fault detection is crucial in modern chemical industries for improved product quality and process safety. In this regard, data-driven (FD) strategy based on independent component analysis (ICA) has gained attention since it improves monitoring by capturing non-gaussian features the data. However, presence measurement noise data degrades performance FD masks important information. To enhance under noisy environment, wavelet-based multi-scale filtering integrated with ICA model to yield a novel Independent (MSICA) strategy. One challenges modeling choose optimum decomposition depth. A scheme parameter estimation at each depth proposed paper achieve this. effectiveness MSICA-based illustrated through three case studies, namely: dynamic multi-variate process, quadruple tank distillation column process. study, MSICA assessed different levels comparing conventional strategies. results indicate that can higher able de-noise capture efficient information from