Intelligent Trend Indices in Detecting Changes of Operating Conditions

Abstract

Temporal reasoning is a very valuable tool to diagnose and control slow processes. Identified trends are also used in data compression and fault diagnosis. Although humans are very good at visually detecting such patterns, for control system software it is a difficult problem including trend extraction and similarity analysis. In this paper, an intelligent trend index is developed from scaled measurements. The scaling is based on monotonously increasing, nonlinear functions, which are generated with generalised norms and moments. The monotonous increase is ensured with constraint handling. Triangular episodes are classified with the trend index and the derivative of it. Severity of the situations is evaluated by a deviation index which takes into account the scaled values of the measurements.