Investigation and Minimization of the Vibrations in High-Precision Manufacturing using Machine Learning

Abstract

The work focuses on fault diagnosis and categorization of rotating equipment parts, using Machine Learning approaches based on the vibration signals pulled from the CWRU Bearing Data Centre at a sampling rate of 48000 samples/sec. The research uses multiple machine learning models, including Support Vector Machines (SVM), 1D Convolution Neural Networks (CNN), 2D CNN, and Long Short-Term Memory (LSTM) networks. In order to minimize overfitting and provide an accurate performance assessment, we used the concept of k-fold cross-validation. Our findings proved the resilience of Machine Learning models for predictive maintenance to detect and identify the faults in the bearings at an early stage. This results in cutting down the frequency of downtime, cost of maintenance, and extending the lifespan of the bearings, so demonstrating the applicability of using Machine Learning in enhancing functionality and monitoring future structures in High-Precision Manufacturing Industries.

Introduction

In high-precision industries, fault diagnosis in rotating machinery is critical to avoid costly downtimes. Predictive maintenance, enhanced through machine learning (ML) and real-time vibration data, shifts strategies from reactive to proactive, improving operational reliability and reducing costs.

Data and Methodology:

The study utilizes the Case Western Reserve University (CWRU) Bearing Dataset, which includes vibration signals under various fault types and severities. The data is preprocessed with noise filtering, normalization, and segmentation, followed by feature extraction (time and frequency domain) and dimensionality reduction (e.g., PCA). Data augmentation and k-fold cross-validation are used to improve model robustness.

Machine Learning Models Used:

• Support Vector Machines (SVM): Efficient with high-dimensional data using kernel functions; fast and computationally light.

• 1D Convolutional Neural Networks (CNN): Best accuracy (95.5% test accuracy), great at learning from raw vibration signals.

• 2D CNNs: Transform vibration data into spectrograms; useful for extracting spatial-frequency features but computationally heavier.

• Long Short-Term Memory (LSTM): Effective with time-sequence patterns; good performance but lower accuracy than CNNs.

Experimental Setup:

Faults were introduced into bearings using electro-discharge machining (EDM). Vibration signals were collected under varying speeds (1730–1797 RPM) and load conditions. Data was used to train and test the models across different operating scenarios.

Results:

• 1D CNN achieved the highest accuracy (95.5%), best F1-score (95.0%), and balanced precision and recall, making it the most effective for fault diagnosis.

• SVM was the most efficient computationally, suitable for real-time deployment with slightly lower accuracy (~92%).

• LSTM showed strong sequential learning capabilities but lagged behind CNNs in performance.

• Overfitting was observed in deep learning models like CNNs and LSTM, indicated by performance gaps between training and test data.

Conclusion

This research systematically evaluated the performance of various machine learning models, including 1D and 2D Convolutional Neural Networks (CNNs), Long Short-Term Memory networks (LSTM), and Support Vector Machines (SVM), in diagnosing faults in rotating machinery based on vibration data. The findings from extensive testing and analysis reveal distinct capabilities and limitations of each model, providing critical insights that could inform their application in industrial settings. 1D CNN emerged as the most effective model in handling complex pattern recognition tasks, achieving the highest accuracy and F1-score among the tested models. This model excelled in extracting meaningful features from raw vibration data, demonstrating substantial robustness and reliability in fault classification. However, it also exhibited signs of overfitting, suggesting a need for improved training strategies or model adjustments to enhance its generalizability. 2D CNN while slightly less effective than its 1D counterpart in overall accuracy, offered advantages in processing data that could be represented in two-dimensional formats, such as spectrograms. This capability makes it particularly useful in scenarios where fault signatures are more discernible in the frequency domain. LSTM provided valuable capabilities in handling data with temporal dependencies, showing promise in applications requiring analysis of sequential data or trends over time. Despite its potential, LSTM\'s performance was slightly lower in standard classification metrics, partly due to its complexity and the challenges associated with training recurrent networks. SVM proved to be highly efficient, particularly in scenarios with limited computational resources. While it did not achieve the highest classification metrics, its speed and lower demand on computational resources make it suitable for real-time applications or as a preliminary fault screening tool. In conclusion, this study demonstrates the potential of machine learning models to revolutionize fault diagnosis in rotating machinery, highlighting specific strengths and applications of different models. By continuing to refine these technologies and address the challenges identified, there is a significant opportunity to enhance predictive maintenance strategies and achieve substantial improvements in industrial operations. The journey from experimental to practical application involves continuous improvement and adaptation, but the path is clear for these advanced diagnostic tools to become integral components of modern industrial systems.

References

[1] Immovilli, F., Bellini, A., Rubini, R., and Tassoni, C. “Diagnosis of bearing faults in induction machines by vibration or current signals: A critical comparison,” IEEE Transactions on Industrial Applications, vol. 46, no. 4, pp. 1350–1359, 2010. [2] Kabiri, P., and Ghaderi, H. “Automobile Independent Fault Detection based on Acoustic Emission Using Wavelet,” Ndt.Net, 2011, pp. 3–4. [3] Patidar, S., and Soni, P. K. “An Overview on Vibration Analysis Techniques for the Diagnosis of Rolling Element Bearing Faults,” International Journal of Engineering Trends and Technology, vol. 4, no. 5, pp. 1804–1809, 2013. [4] Chaudhari, Y. K., Gaikwad, J. A., and Kulkarni, J. V. “Vibration analysis for bearing fault detection in electrical motors,” presented at the 1st International Conference on Networks & Soft Computing, 2014, pp. 146–150. [5] Srivastava, N., Hinton, G., Krizhevsky, A., Sutskever, I., and Salakhutdinov, R. “Dropout: a simple way to prevent neural networks from overfitting.” Journal of Machine Learning Research, vol. 15, no. 1, pp. 1929–1958, 2014. [6] Gupta, V. K., et al. “Ball Bearing Fault Diagnosis using Supervised and Unsupervised Machine Learning Methods.” International Journal of Acoustics and Vibration, vol. 20, (2015), 482–485. [7] Kulkarni, S., and Bewoor, A. “Vibration-based condition assessment of ball bearing with distributed defects,” Journal of Measurement Engineering, vol. 4, no. 2, pp. 87–94, 2016. [8] Caesarendra, W., and Tjahjowidodo, T. “A review of feature extraction methods in vibration-based condition monitoring and its application for degradation trend estimation of low-speed slew bearing.” Machines 5, no. 4 (2017): 21. [9] Ben Abid, F., and Braham, A. “Advanced signal processing techniques for bearing fault detection in induction motors,” presented at the 2018 15th International Multi-Conference on Systems, Signals & Devices (SSD), 2018, pp. 882–887. [10] Elforjani, M., and Shanbr, S. “Prognosis of Bearing Acoustic Emission Signals Using Supervised Machine Learning.” IEEE Transactions on Industrial Electronics, vol. 65, no. 7, (2018), 5864–5871. [11] Osman, H. A., Salman, A. A., and Fawzy, F. M. “Vibration Signature of Roller Bearing’s Faults,” European Scientific Journal (ESJ), vol. 15, no. 12, 2019. [12] Eren, L., et al. “A generic intelligent bearing fault diagnosis system using compact adaptive 1D CNN classifier.” Journal of Signal Processing Systems, vol. 91, (2019), 179–189. [13] Wescoat, E., Mears, L., Goodnough, J., and Sims, J. “Frequency energy analysis in detecting rolling bearing faults,” Procedia Manufacturing, vol. 48, pp. 980–991, 2020. [14] Roy, S. S., Dey, S., and Chatterjee, S. “Autocorrelation aided random forest classifier-based bearing fault detection framework.” IEEE Sensors Journal, vol. 20, no. 18, pp. 10792–10800, 2020. [15] Liu, N., Liu, B., and Xi, C. “Fault diagnosis method of rolling bearing based on the multiple features of LMD and random forest,” In IOP Conference Series: Materials Science and Engineering, vol. 892, no. 1, 012068, 2020. [16] Kong, X., Mao, G., Wang, Q., Ma, H., and Yang, W. “A multi-ensemble method based on deep auto-encoders for fault diagnosis of rolling bearings.” Measurement, vol. 151, 107132, 2020. [17] Goyal, D., Dhami, S. S., and Pabla, B. S. “Non-Contact Fault Diagnosis of Bearings in Machine Learning Environment.” IEEE Sensors Journal, vol. 20, no. 9, (2020), 4816–4823. [18] Wang, Z., Ma, H., Chen, H., Yan, B., and Chu, X. “Performance degradation assessment of rolling bearing based on convolutional neural network and deep long-short term memory network.” International Journal of Production Research, vol. 58, no. 13, (2020), 3931–3943. [19] Hariharan, V., Thangavel, P., Rajeshkumar, G., and Deepa, D. “Investigations of Antifriction bearing defects using Vibration Signatures,” IOP Conference Series: Materials Science and Engineering, vol. 1084, no. 1, 012126, 2021. [20] Barcelos, A. S., and Marques Cardoso, A. J. “Current-based bearing fault diagnosis using deep learning algorithms,” Energies, vol. 14, no. 9, 2509, 2021. [21] Schwendemann, S., Amjad, Z., and Sikora, A. “A survey of machine-learning techniques for condition monitoring and predictive maintenance of bearings in grinding machines,” Computers in Industry, vol. 125, 103380, 2021. [22] Barai, V., Ramteke, S. M., Dhanalkotwar, V., Nagmote, Y., Shende, S., and Deshmukh, D. “Bearing fault diagnosis using signal processing and machine learning techniques: A review.” In IOP Conference Series: Materials Science and Engineering, vol. 1259, no. 1, p. 12–34, 2022. [23] Amirat, H. Y., et al. “Bearing Fault Event-Triggered Diagnosis Using a Variational Mode Decomposition-Based Machine Learning Approach.” IEEE Transactions on Energy Conversion, vol. 37, no. 1, (2022), 466–474. [24] Kulkarni, A. M., et al. “Real Time Cloud based Fault Detection and Alert for Antenna Array Using CNN.” IETE Journal of Research (2023), 1–12.

Copyright

Copyright © 2025 K Vardhrajan, Dr. Pradeep Kumar Sharma, Mr. Surendra Kumar. This is an open access article distributed under the Creative Commons Attribution License, which permits unrestricted use, distribution, and reproduction in any medium, provided the original work is properly cited.