Vibration-Based Fault Diagnosis with Low Delay

This is the code for the paper entitled "A Hybrid Method for Condition Monitoring and Fault Diagnosis of Rolling Bearings With Low System Delay", published in the IEEE Transactions on Instrumentation and Measurement, August 2022.
Authors: Sulaiman Aburakhia, Ryan Myers, and Abdallah Shami.
Organization: The Optimized Computing and Communications (OC2) Lab, ECE Department, Western University, London, Canada.

The system delay $\tau_d$ of a vibration-based condition monitoring system can be defined as the time the system takes to acquire input vibration segment and classify or predict the operational condition of the current state of the equipment. In vibration-based monitoring, the current state $S_c$ is represented by the input segment $v_{in}$ of the generated vibration signal as shown in the figure below. Accordingly, the system delay is the sum of the time duration of input segment $T_{v_{in}}$ and the online processing time $T_p$ .

The time duration of the input segment depends on number of data points in the segment.
The online processing time is algorithm-dependent; it involves two tasks, feature extraction (including pre-processing) and condition prediction/classification.

Hence, online processing time can be generally viewed as a function of the number of data points in the input segment, the size of extracted features, and available computing resources. Thus, designing a condition monitoring system with low system delay involves three main requirements:

Extracting features of high sensitivity to fault-related transients to improve system accuracy.
Extracting features of small size.
Utilizing input vibration segments of relatively short time duration or equivalently, of small number of data points.

Accordingly, the paper utilizes wavelet decomposition and Fourier analysis, and proposes a hybrid method to fulfill the aforementioned requirements and extract a small number of highly discriminative features from short-duration vibration signals. The first step involves decomposing the input vibration segment using $k$-level Wavelet Packet Transform (WPT). Consequently, $2^k$ elementary waveforms of lower and higher frequency sub-bands are reconstructed from individual wavelet coefficients. In the second step, Fast Fourier Transform (FFT) is applied to the resultant waveforms to obtain the spectral contents of each waveform. Hence, a feature vector of size $S=1×(m×2^k)$ can be reconstructed by utilizing the first $m$ dominant frequency components in the spectrum of each waveform as illustrated in the figure below. For more details, please refer to the paper.

The Performance of the proposed method is evaluated on the Case Western Reserve University (CWRU) bearing dataset, the Paderborn University (PU) bearing dataset, and the University of Ottawa (uOttawa) bearing dataset. These datasets are selected to simulate various practical situations regarding defect types, rotational speed conditions, and data sampling rate.

Codes and Datasets:

A separate Jupyter notebook is provided for each of the three datasets. Functions for processing .mat vibration files and creating training/testing datasets are included in the notebooks as well.

Datasets:

the Case Western Reserve University (CWRU) bearing dataset
the Paderborn University (PU) bearing dataset
the University of Ottawa (uOttawa) bearing dataset

Python Codes:

Code for the CWRU dataset
Code for the PU dataset
Code for the uOttawa dataset

Contact Information

For all inquiries or collaboration opportunities please contact:

Email : [email protected] or [email protected]
Github: SulAburakhia or Western OC2 Lab
Google Scholar: OC2 Lab; Sulaiman Aburakhia

Citation

If you find this repository useful in your research, please cite as:

S. A. Aburakhia, R. Myers and A. Shami, "A Hybrid Method for Condition Monitoring and Fault Diagnosis of Rolling Bearings With Low System Delay," in IEEE Transactions on Instrumentation and Measurement, vol. 71, pp. 1-13, 2022, Art no. 3519913, doi: 10.1109/TIM.2022.3198477.

@ARTICLE{9855510,
  author={Aburakhia, Sulaiman A. and Myers, Ryan and Shami, Abdallah},
  journal={IEEE Transactions on Instrumentation and Measurement}, 
  title={A Hybrid Method for Condition Monitoring and Fault Diagnosis of Rolling Bearings With Low System Delay}, 
  year={2022},
  volume={71},
  number={},
  pages={1-13},
  doi={10.1109/TIM.2022.3198477}}

Publication

Pre-print is available here.

tdl77 / vibration-based-fault-diagnosis-with-low-delay Goto Github PK