Logo
ScienceStack
Table of Contents
Fetching ...
Sign In

Statistical Batch-Based Bearing Fault Detection

Victoria Jorry, Zina-Sabrina Duma, Tuomas Sihvonen, Satu-Pia Reinikainen, Lassi Roininen

TL;DR

Rolling-element bearing faults (ball, inner, outer race) pose maintenance and safety challenges under time-varying conditions. The authors present a batch-based MSPC framework that extracts Fourier-transform features from fixed-length windows, builds PCA-based health models, and monitors $T^2$ and $SPEx$ for fault detection without requiring faulty training data. They optimize batch length, number of FT components, and feature subsets via an Augmented Lagrangian GA with backward variable selection, achieving robust fault detection across locations and loads on the CWRU bearing dataset, with near-zero false alarms in combined-load scenarios. The approach offers fast calibration, scalable real-time monitoring, and potential industrial applicability for proactive bearing maintenance.

Abstract

In the domain of rotating machinery, bearings are vulnerable to different mechanical faults, including ball, inner, and outer race faults. Various techniques can be used in condition-based monitoring, from classical signal analysis to deep learning methods. Based on the complex working conditions of rotary machines, multivariate statistical process control charts such as Hotelling's $T^2$ and Squared Prediction Error are useful for providing early warnings. However, these methods are rarely applied to condition monitoring of rotating machinery due to the univariate nature of the datasets. In the present paper, we propose a multivariate statistical process control-based fault detection method that utilizes multivariate data composed of Fourier transform features extracted for fixed-time batches. Our approach makes use of the multidimensional nature of Fourier transform characteristics, which record more detailed information about the machine's status, in an effort to enhance early defect detection and diagnosis. Experiments with varying vibration measurement locations (Fan End, Drive End), fault types (ball, inner, and outer race faults), and motor loads (0-3 horsepower) are used to validate the suggested approach. The outcomes illustrate our method's effectiveness in fault detection and point to possible broader uses in industrial maintenance.

Statistical Batch-Based Bearing Fault Detection

TL;DR

Rolling-element bearing faults (ball, inner, outer race) pose maintenance and safety challenges under time-varying conditions. The authors present a batch-based MSPC framework that extracts Fourier-transform features from fixed-length windows, builds PCA-based health models, and monitors and for fault detection without requiring faulty training data. They optimize batch length, number of FT components, and feature subsets via an Augmented Lagrangian GA with backward variable selection, achieving robust fault detection across locations and loads on the CWRU bearing dataset, with near-zero false alarms in combined-load scenarios. The approach offers fast calibration, scalable real-time monitoring, and potential industrial applicability for proactive bearing maintenance.

Abstract

In the domain of rotating machinery, bearings are vulnerable to different mechanical faults, including ball, inner, and outer race faults. Various techniques can be used in condition-based monitoring, from classical signal analysis to deep learning methods. Based on the complex working conditions of rotary machines, multivariate statistical process control charts such as Hotelling's and Squared Prediction Error are useful for providing early warnings. However, these methods are rarely applied to condition monitoring of rotating machinery due to the univariate nature of the datasets. In the present paper, we propose a multivariate statistical process control-based fault detection method that utilizes multivariate data composed of Fourier transform features extracted for fixed-time batches. Our approach makes use of the multidimensional nature of Fourier transform characteristics, which record more detailed information about the machine's status, in an effort to enhance early defect detection and diagnosis. Experiments with varying vibration measurement locations (Fan End, Drive End), fault types (ball, inner, and outer race faults), and motor loads (0-3 horsepower) are used to validate the suggested approach. The outcomes illustrate our method's effectiveness in fault detection and point to possible broader uses in industrial maintenance.
Paper Structure (16 sections, 10 equations, 10 figures, 2 tables)

This paper contains 16 sections, 10 equations, 10 figures, 2 tables.

Table of Contents

  1. Introduction
  2. Materials and methods
  3. Dataset description
  4. Methods
  5. Batching procedure and Fourier transform
  6. Feature extraction
  7. Principal Component Analysis (PCA)
  8. Variable, batch and feature optimization
  9. Results and discussion
  10. Fourier transform of vibrational data
  11. Individual HP PCA models and control charts
  12. Combined-HP PCA models and control charts
  13. Model optimization and variable selection
  14. Fan End (FE) models and combined Drive End-Fan End (DE-FE) models
  15. Comparison with other fault identification methods
  16. ...and 1 more sections

Figures (10)

  • Figure 1: Zoomed-in CWRU raw vibration signals at a motor load of 0 hp and speed of 1797 rpm.
  • Figure 2: Workflow for fault detection with window-based FT MSPC.
  • Figure 3: Backwards selection workflow: Variable optimization for fault identification.
  • Figure 4: FFT decomposition for DE data for normal(healthy) bearing operation. In (a), the initial batch length is 1000 timestamps, whereas in (b) 5180 timestamps were considered. The timelines have been shortened for visualisation purposes.
  • Figure 5: FFT decomposition for the inner race DE fault data. In (a), the initial batch length is 1000 timestamps, whereas in (b), 5180 timestamps were considered. The timelines have been shortened for visualisation purposes.
  • ...and 5 more figures