Enhancing the noise tolerance of fault diagnosis system using the modified adaptive boosting algorithm

• A major problem associated with this method is the fact that it is not strong enough and sometimes fails in noisy environments.
• The aim of the study was to enhancing the noise tolerance of a fault diagnosis system using the modified adaptive boosting algorithm.
• In this paper, to eliminate this drawback, the weighting system of the typical AdaBoost algorithm was modified.

Recent investigations in the field of fault diagnosis have increasingly used ensemble learning techniques due to their exceptional performance in dealing with high-dimensionality, small sample size, and complex data structure. An ensemble of ANNs can enhance the generalizability and reliability of a single ANN system, through training the ANNs for a given assignment and incorporating the outcomes. One of the conventional and widely recognized ensemble methods is the adaptive boosting (AdaBoost). A major problem associated with this method is the fact that it is not strong enough and sometimes fails in noisy environments. In this paper, to eliminate this drawback, the weighting system of the typical AdaBoost algorithm was modified. This paper applies the new boostig (MadaBoost) algorithm for the fault diagnosis of a chemical plant by using an ensemble of neural networks, with the objective to improve the noise tolerance of fault diagnosis system using the modified adaptive boosting algorithm.