An online algorithm with confidence for real-time fault detection

作者：Wang Huazhen^*; Lin Chengde; Yang Fan; Zhuang Jinfa

来源：Journal of Information and Computational Science, 2009, 6(1): 305-313.

• Algorithmic process
• Compression model
• Conformal predictor
• Large-scale industrial process
• On-line algorithms
• Online learning
• Random forest
• Real time fault detection
• Tennessee eastman process
• Transductive inference

Fault detection in large-scale industrial process should work online as well as providing confidence-level of prediction. Conformal predictor is a compression model framework based on on-line learning and transductive inference. It predicts individual cases coupled with valid confidence. However, it has an obvious disadvantage that algorithmic process is computational expensive. In this paper, conformal predictor is extended to hybrid-compression conformal predictor to upgrade the computational efficiency without losing too much information. A case study of Tennessee Eastman Process is provided to illustrate the computational and predictive efficiency of the proposed method.

• 出版日期2009
• 单位厦门大学