Multiscale Detection of Chemical Process using Improved Distributed CCA-Wavelet Approach

In the last two decades, industrial process systems are more complicated and dynamic due to the rapid advancement of plant automation and computerized-aided sensor systems. Process monitoring has been a crucial research field in the industry to improve process quality and plant safety. Traditional detection approaches are single-scale that cannot deal with the dynamic multi-dimensional correlated data generated by complex automated industrial processes. Multiscale detection plays a vital role in the monitoring of industrial systems. Therefore, an improved multiscale distributed canonical correlation analysis (MD-CCA) detection framework is proposed to improve detection tendency in the industrial process system. This framework integrates the data-driven multiscale detection approaches based on wavelet transforms (WT) and distributed canonical correlation analysis (D-CCA). The proposed and existing frameworks' effectiveness is evaluated and differentiated using a continuous stirred tank reactor (CSTR) system as an application case study. The results indicate that the proposed (MD-CCA) framework detects abnormalities and faults more efficiently and robustly than the existing D-CCA approach. This concludes that the proposed method is effective and efficient.