Abstract:The failure of rolling bearings causes performance deterioration and even unplanned downtime of rotating machinery. Intelligent classifiers based on machine learning algorithms have been well-established recently to recognize different types of rolling bearing failure. Usually, the faulty data on field is incomplete in fault types and lacks in quantity. Conventional classification algorithms may fail to detect a new class and identify one class wrongly as the other classes. In this paper, extreme function theory is introduced to detect new classes and perform fault classification under the assumption that limited class and small datasets are available for training. A Gaussian process (GP) model is created from the training data containing only one class. The probability density of any observations can be obtained from that model. According to the extreme value theory, that Probability Density Function (PDF) will have a corresponding Extreme Value (EV) distribution belonging to an extreme distribution. Choosing a proper confidence interval, a threshold can be calculated from a fitted Cumulative Distribution Function (CDF). The outliers of the EVs suggest a new class not seen in the training set, detecting thus potential faults, by also keeping a low ratio of false alarms. With the GP models trained from different classes, different thresholds indicating whether the observation belongs to the trained class are obtained. The multiclass classification can be realised in such a one-against-all architecture. The experimental data of rolling bearings in normal operation and three different faulty conditions are used in this paper to verify the effectiveness of extreme function theory for fault detection and classification. This approach is also compared with Artificial Neural Network in terms of false identification ratio.

Wear Particle Classification Using Genetic Programming Evolved Features

Genetic programming-based classification of ferrograph wear particles

Ferrography Wear Particle Pattern Recognition Based on Support Vector Machine

Ferrograph Analysis with Improved Particle Segmentation and Classification Methods

Integrated Model of BP Neural Network and CNN Algorithm for Automatic Wear Debris Classification

Application of Genetic Programming in the Identification of Tool Wear

Fault Detection and Classification of Rolling Bearings Using Extreme Function Theory

Shape classification of wear particles by image boundary analysis using machine learning algorithms

A Hybrid Convolutional Neural Network for Intelligent Wear Particle Classification

Wear Particles Pattern Recognition Based On Neural Network

A Prototype of On-Line Extraction and Three-Dimensional Characterisation of Wear Particle Features from Video Sequence

Online wear characterisation of rolling element bearing using wear particle morphological features

Tool wear state prediction based on feature-based transfer learning

A Hybrid Search-Tree Discriminant Technique for Multivariate Wear Debris Classification

Ferrography Wear Particles Image Recognition Based on Extreme Learning Machine

Optimized CNN Model for Identifying Similar 3D Wear Particles in Few Samples

WP-DRnet: A Novel Wear Particle Detection and Recognition Network for Automatic Ferrograph Image Analysis

Tool wear state recognition based on GWO–SVM with feature selection of genetic algorithm

Analysis of typic Wear Particles based on Principal Component Classification

Shape and texture features in the automated classification of adhesive and abrasive wear particles

Application of Improved Genetic Programming for Feature Extraction in the Evaluation of Bearing Performance Degradation