Vibration-Based Fault Diagnosis of Slurry Pumps Using the Neighborhood Rough Set Model

Abstract

Rough set has been widely used as a method of feature selection in fault diagnosis. The neighborhood rough set model can deal with both nominal and numerical features, but selecting the neighborhood size for its application may be a challenge. In this paper, we illustrate that using a single neighborhood size for all features may overestimate or underestimate a feature’s degree of dependency. The neighborhood rough set model is then modified by setting different neighborhood sizes for different features. The modified model is applied to fault diagnosis of slurry pump impellers. The chosen feature subsets generated by the modified rough set model can be physically explained by the corresponding flow patterns and generate higher classification accuracy than the original feature subsets and the feature subsets generated by the original rough set model.