プラント起動時における経験的診断ロジックの動的時間伸縮法による設計法

概要

論文の詳細を見る
A novel design method of empirical if-then rules for diagnosis of plant start-up sequences is proposed. The rules dealt with in this paper take form of "If variable Y is out of the determined range when variable X becomes X_0, then raise alarm", which is one of the styles of rules commonly implemented in practical plant control logic. When sufficient number of normal and abnormal data is available, one can build empirical if-then rules for early detection of abnormal start-up sequences. Conventionally the rules are built by expert engineers based on their experiences, and inevitably in subjective way. Present paper proposes a design method which can objectively give an optimal rule. The first step of the proposed method is 'alignment' of data profiles using dynamic time warping. The second step is calculations of correlation ratios of data to find the best condition to distinguish normal and abnormal start-up sequences. A simulation study of a natural gas fueled plant is used as an example of applications of the proposed method.
2010-07-25