Complex dynamical system fault diagnosis based on multiple ANFIS using independent component

Abstract

In this paper, an online fault diagnosis for a complex dynamical systems integrating adaptive neuro-fuzzy inference system (ANFIS) and using independent component analysis (ICA) for feature extracting is presented. In this approach, using ICA provide salient features selected from raw measured data sets. Subsequently, the most superior extracted features are fed into multiple ANFIS in order to identify different abnormal cases. Using multiple ANFIS units each dedicated to one fault, significantly reduces the scale and complexity of the system and speeds up the training of the network. This data-driven method is quite simple fast and expandable. Chemical dynamic processes are so complex that they are presently ahead of theoretical methods from fundamental physical standpoint. Distillation column is used as a complex benchmark problem to evaluate the proposed algorithm. Experimental results from this simulated nonlinear MIMO dynamical system demonstrated the effectiveness of the method. This procedure is applicable to a variety of industrial applications in which continuous on-line monitoring and diagnosis are necessary.