Pattern Recognition of Bearing Faults using Smoother Statistical Features

TL;DR

This paper presents a pattern recognition method for bearing fault diagnosis that uses smoothed statistical features from vibration data to improve diagnostic accuracy by reducing randomness effects.

Contribution

It introduces a novel feature smoothing technique to enhance the reliability of pattern recognition in bearing fault diagnosis.

Findings

Smoothing of features improves fault classification accuracy.

The method effectively reduces the impact of vibration randomness.

Enhanced diagnostic performance demonstrated on test rig data.

Abstract

A pattern recognition (PR) based diagnostic scheme is presented to identify bearing faults, using time domain features. Vibration data is acquired from faulty bearings using a test rig. The features are extracted from the data, and processed prior to utilize in the PR process. The processing involves smoothing of feature distributions. This reduces the undesired impact of vibration randomness on the PR process, and thus enhances the diagnostic accuracy of the model.