An improved method for fault diagnosis of a batch process by use of pattern-recognition technique.

DOI Web Site Web Site 1 Citations Open Access

Bibliographic Information

Other Title
  • パターン認識を利用したバッチプロセスの異常診断法の改良
  • パターン ニンシキ オ リヨウシタ バッチ プロセス ノ イジョウ シンダンホ

Search this article

Description

In the fault diagnosis of a batch process utilizing the pattern-recognition technique, the region of abnormal patterns corresponding to each causal event is defined by analyzing the accumulated operation data and the origin of failure currently faced is identified by matching the pattern observed in the process with the region. Modification of the region by the use of new data (learning) plays an important role in this method for fault diagnosis. Enlargement of the region makes it possible to prevent the wrong diagnosis, but causes the detection of plural candidates for the origin of failure. In the present paper, a definition of the region and a method for the learning are proposed to reduce the number of detected candidates by use of the degree of distinguishability from the abnormal patterns due to the different causal events. Experimental results confirmed that the repetition of learning can reduce not only the frequency of wrong diagnosis but also the number of candidates.

Journal

Citations (1)*help

See more

Keywords

Details 詳細情報について

Report a problem

Back to top