An Integrated Pattern Query Method to Assist Plant Monitoring

Live plant behavior is usually recorded and saved as multivariate time series data which can be explored to assist process engineers in process monitoring, abnormality/fault detection and diagnosis. For example, patterns that indicate a trend or a fault might repeat itself over the whole operating period. In addition, faced with abundant process data, process engineers are challenged to efficiently and effectively establish connections between specific variables who share similar patterns. In current work, we demonstrate the adoption of an integrated pattern query framework which combines PCA similar factor and a fast similarity search algorithm to overcome such a challenge. The efficacy and efficiency of proposed framework is illustrated by both simulation and industrial case studies. Such results demonstrate the successful application of data analytics to cope with real problems in chemical manufacturing within the Digital Transformation era.