A novel key performance indicator oriented process monitoring method based on multiple information extraction and support vector data description

As a core part of modern chemical plants, key performance indicator oriented process monitoring and fault diagnosis systems have gradually made great contributions to guaranteeing process safety, improving product quality, and ensuring system reliability, which recently have received extensive attention and become one of the hot spots both in academic research and industrial application fields. Different from previous methods, a novel key performance indicator oriented process monitoring method is proposed in this paper, which fully mines and utilizes important time feature information hidden in the process data while considering the local process information. Firstly, a group of representative process variables with maximum key performance indicator information are selected by the maximal information coefficient algorithm, and local information is extracted. Then, observed value, accumulated error, and change rate information are further extracted from the representative process variables and expanded into multiple information blocks, which contain both local process and hidden time feature information. After that, the support vector data description model is established to monitor each information block, and the Bayesian inference is employed to fuse the final monitoring results to obtain a new monitoring index. Finally, the performance and effectiveness of the proposed method is validated by conducting a simulation on Tennessee Eastman process.