Fault detection through evolving fuzzy cloud-based model

Abstract

An evolving fuzzy model for fault detection is presented in this paper. The method is based on simplified, non-parametric fuzzy model named AnYa. The novelty in this paper is the partial density estimation where only the most influential components are used. The proposed method is tested on simulated data of Tennessee Eastman Process model and furthermore, the results are compared with well established fault detection methods, i.e. PCA (Partial Component Analysis), ICA (Independent Component Analysis), and FDA (Fisher Discriminant Analysis). The results show that the proposed method is capable of detecting different fault types with very high accuracy.