プラント異常診断支援システムのための定量モデルによる知識表現

概要

論文の詳細を見る
This paper proposes a quantitative-model-based approach to the knowledge representation for the process diagnosis system to support plant operation. Observed operators' processing may justifiably recommend the usage of the quantitative process model. The process model is basically made with the knowledge in chemical engineering, and practically modified with time to follow the varying state of the plant process. On the basis of the process model, the representative term "balance-measure" is newly introduced. The terms in each model expression equivalently explain the balanced situation in the model, and are availed to evaluate the process situation. Plant feature prior to the present is described in terms of the balance-measure term from the data already gathered and compiled as the models to be referred in the form of the set of the term. The present process state is also described in the term. After comparison between the present state and the references, estimated difference between them suggests the faulty changes in the plant process. The process state variable is defined to compose of three elements ; the theoretical value in the model, the indicated measure, and the deviation between the prior two. The diagnosis is performed by inferring the value of the deviation in each model.
1990-03-01