A TSK-Type Recurrent Neuro-Fuzzy Systems for Fault Prognosis

Abstract

As a result from the demanding of process safety, reliability and environmental constraints, a called of fault detection and diagnosis system become more and more important. In this article some basic aspects of TSK (Takigi Sugeno Kang) neuro-fuzzy techniques for the prognosis and diagnosis of manufacturing systems are presented. In particular, a neuro-fuzzy model that can be used for the identification and the simulation of faults prognosis models is described. The presented model is motivated by a cooperative neuro-fuzzy approach based on a vectorized recurrent neural network architecture. The neuro-fuzzy architecture maps the residuals into two classes: a one of fixed direction residuals and another one of faults belonging to rotary kiln

A review on prognostic techniques for non-stationary and non-linear rotating systems

“…Other applications in the use of NF systems for prognostics can be found in Refs. [179,[181][182][183].…”

“…This technique has an easy rule-based design with linguistic modelling features, learning and fault-tolerance ability, and parallel processing functions. NF systems are nonlinear in nature and can be applied to complex systems [179]. A limitation of NF systems is that implementation relies heavily on prior knowledge of a system and the availability of empirical data.…”

“…Overcomes the disadvantages of fuzzy system by integrating it with neural network to train the fuzzy structure and parameters [163,178,213] Very reliable and robust prognostic technique with high forecasting accuracy [163] Fast and accurately capture and model the system dynamic behaviour so that it can be further utilized to perform prognosis of machine health [163,181] The most promising flexible-model technique in areas with high uncertainty and complexity [163,179,213] Model design based on linguistic rules making it easy to comprehend [179,213] Can apply to deal with non-stationary operating condition [178,214] Inherent non-linear nature [179,213] Parallel processing and fault-tolerance abilities [179,213] Short prediction horizons [12] Require considerable amount of historical data with the corresponding degradation states and prior knowledge of the system, or else it will suffer from poor prediction accuracy in case of small training dataset or with fast dynamic fluctuations, such as during the chipping of gear tooth surface material or just prior to gear failure [12,179,183,213] Unable to update the system states using the updated new data [181] With fixed reasoning structures but without sufficient adaptive capability to accommodate time-varying dynamic effects [180] …”

Multi-Sensor Condition Monitoring of Bearings Using Support Vector Machines

“…• Overcomes the disadvantages of fuzzy system by integrating it with neural network to train the fuzzy structure and parameters [165][166][167] • Very reliable and robust prognostic technique with high forecasting accuracy [166] • Fast and accurately capture and model the system dynamic behaviour so that it can be further utilized to perform prognosis of machine health [166,168] • The most promising flexible-model technique in areas with high uncertainty and complexity [165,166,169] • Model design based on linguistic rules making it easy to comprehend [165,169] • Can apply to deal with non-stationary operating condition [167,170] • Inherent non-linear nature [165,169] • Short prediction horizons [65] • • Unable to update the system states in real time using the updated online new data [168] • With fixed reasoning structures but without sufficient adaptive capability to accommodate timevarying dynamic effects [172] • Parallel processing and fault-tolerance abilities [165,169] …”

“…This technique has an easy rule-based design with linguistic modelling features, learning and fault-tolerance ability, and parallel processing functions. NF systems are nonlinear in nature and can be applied to complex systems [169]. A limitation of NF systems is that implementation relies heavily on prior knowledge of a system and the availability of empirical data.…”