Decision Tree Based Qualitative Analysis of Operating Regimes in Industrial Production Processes

Abstract

The qualitative analysis of complex process systems is an important task at the design of control and process monitoring algorithms. Qualitative models require interpretable description of the operating regimes of the process. This work shows a novel approach to discover and isolate operating regimes of process systems based on process models, time series analysis, and decision tree induction technique. The novelty of this approach is the application of time series segmentation algorithms to detect the homogeneous periods of the operation. Advanced sequence alignment algorithm used in bioinformatics is applied for the calculation of the similarity of the process trends described by qualitative variables. Decision tree induction is applied for the transformation of this hidden knowledge into easily interpretable rule base to represent the operation regions of the process. The whole methodology is applied to detect operating regimes of an industrial fixed bed tube reactor.