โก Quick Summary
This study introduces a Transformer-based Variational Autoencoder (T-VAE) designed to generate synthetic training data for anomaly detection in high-speed spindle motors. By augmenting the training dataset, the model significantly improved detection accuracy, achieving 98.07% pass detection for normal samples and 97.99% fail detection for faulty samples.
๐ Key Details
- ๐ Dataset: 10,000 real samples (5,000 normal and 5,000 faulty) from three spindle motors
- โ๏ธ Technology: Transformer-based Variational Autoencoder (T-VAE)
- ๐ Performance: CNN-LSTM classifier with and without data augmentation
- ๐ Evaluation: Independent dataset of 50,000 sequences (25,000 normal and 25,000 faulty)
๐ Key Takeaways
- ๐ง Vibration analysis is crucial for detecting mechanical anomalies in spindle motors.
- ๐ Data scarcity limits the effectiveness of deep learning models in anomaly detection.
- ๐ค T-VAE generates realistic synthetic vibration data to augment training datasets.
- ๐ Performance boost: Augmentation led to a significant increase in detection accuracy.
- ๐ Cross-spindle evaluation demonstrated the model’s robustness across different spindle motors.
- ๐ Direct application in predictive maintenance systems for manufacturing environments.
๐ Background
In the realm of high-speed spindle motors, operating at speeds exceeding 10,000 rpm, vibration analysis plays a pivotal role in identifying mechanical anomalies. However, the challenge of data scarcity and imbalance, particularly for rare fault conditions, hampers the performance of deep learning models. This study addresses these challenges by proposing a novel approach to generate synthetic training data.
๐๏ธ Study
The research involved the collection of 10,000 real vibration samples from three spindle motors, comprising 5,000 normal and 5,000 faulty samples. The authors developed the Transformer-based Variational Autoencoder (T-VAE), which utilizes positional encoding and multi-head self-attention mechanisms to effectively capture long-range temporal dependencies in multivariate time-series data. A KL annealing strategy was also implemented to enhance training stability.
๐ Results
The results were promising: the CNN-LSTM classifier achieved a 95.73% pass detection rate for normal samples and 81.40% fail detection rate for faulty samples without data augmentation. After augmenting the training dataset with synthetic samples generated by the T-VAE, the performance improved dramatically, reaching 98.07% pass detection for normal data and 97.99% fail detection for faulty data.
๐ Impact and Implications
The findings of this study have significant implications for the field of predictive maintenance in manufacturing. By effectively addressing the data scarcity problem, the T-VAE model enhances the accuracy of anomaly detection in spindle motor vibration signals. This advancement could lead to more reliable maintenance strategies, ultimately improving operational efficiency and reducing downtime in industrial settings.
๐ฎ Conclusion
This study highlights the transformative potential of using a Transformer-based Variational Autoencoder for generating synthetic training data in anomaly detection. The significant improvements in detection accuracy underscore the importance of innovative approaches to tackle data scarcity challenges. As industries continue to embrace predictive maintenance technologies, the integration of such models will likely play a crucial role in enhancing operational reliability and efficiency.
๐ฌ Your comments
What are your thoughts on the use of synthetic data generation in anomaly detection? We would love to hear your insights! ๐ฌ Leave your comments below or connect with us on social media:
A Transformer-Based Variational Autoencoder for Training Data Generation in Spindle Motor Vibration-Based Anomaly Detection.
Abstract
In high-speed spindle motors operating above 10,000 rpm, vibration analysis is essential for detecting mechanical anomalies. However, data scarcity and imbalance, especially for rare fault conditions, limit the performance of deep learning-based anomaly detection models. In this study, we define sample scarcity as the limited availability of real labeled vibration sequences for model training, i.e., only 5000 normal and 5000 faulty samples collected from three spindle motors (10,000 real samples in total). We propose a Transformer-based Variational Autoencoder (T-VAE) to generate realistic triaxial acceleration sequences for spindle motor health monitoring. The model integrates positional encoding and multi-head self-attention to capture long-range temporal dependencies in multivariate time-series data, and applies a KL annealing strategy to improve training stability. Using 5000 normal and 5000 faulty vibration samples collected from three spindle motors, the model generates 100,000 synthetic samples per class, which are used to augment training for a downstream CNN-LSTM classifier. Without augmentation, the classifier achieved 95.73% pass detection on normal samples and 81.40% fail detection on faulty samples. After augmentation with Transformer-VAE, performance increased to 98.07% pass detection for normal data and 97.99% fail detection for faulty data. For prediction, we evaluate on an independent dataset of 25,000 normal and 25,000 faulty sequences obtained from eleven different spindle motors not used in training (cross-spindle). The results demonstrate that the T-VAE effectively alleviates the data scarcity problem and significantly improves anomaly detection accuracy for high-speed spindle motor vibration signals. This approach can be directly applied to predictive maintenance systems in real-world manufacturing environments.
Author: [‘Kim J’, ‘Hwang Y’]
Journal: Sensors (Basel)
Citation: Kim J and Hwang Y. A Transformer-Based Variational Autoencoder for Training Data Generation in Spindle Motor Vibration-Based Anomaly Detection. A Transformer-Based Variational Autoencoder for Training Data Generation in Spindle Motor Vibration-Based Anomaly Detection. 2026; 26:(unknown pages). doi: 10.3390/s26072176