Intelligent Fault Diagnosis for Imbalanced Data

Intelligent fault diagnosis for imbalanced data addresses the pervasive challenge in condition monitoring where fault instances are scarce relative to normal operating samples. Conventional machine-learning and deep-learning approaches can neglect rare fault events, leading to poor detection of critical failures. Recent work has focused on developing adaptive resampling methods, cost-sensitive algorithms and representation learning techniques that enhance minority-class recognition without sacrificing general performance. These solutions span classical classifiers augmented by hybrid over- and under-sampling, convolutional and recurrent neural networks with adaptive loss functions, generative models for synthetic data creation and meta-learning frameworks for few-shot adaptation. Practical applications include rotating machinery, electrical machines, high-speed rail traction devices and aerospace systems, where early and accurate fault detection under data imbalance yields significant gains in safety, reliability and maintenance efficiency.

Research from Nature Portfolio

No recent Nature Portfolio content available.

Intelligent Fault Diagnosis for Imbalanced Data publication trend

The graph below shows the total number of articles in intelligent fault diagnosis for imbalanced data across all publications each year (not limited to Nature Index journals).

Technical terms

Imbalanced data: A dataset in which one class (usually normal operating samples) significantly outnumbers others (rare fault events), leading to biased model training.

Synthetic minority oversampling technique (SMOTE): An algorithm that generates synthetic examples for the minority class by interpolating between existing samples to rebalance class distribution.

Tomek link: A sample-cleaning method that identifies and removes overlapping majority-class samples adjacent to minority samples, improving class separation.

Meta-learning: A framework in which models learn how to learn, enabling rapid adaptation to new tasks with limited data by leveraging prior experience.

Few-shot learning: The ability of a model to generalise and recognise new classes from only a handful of labelled examples, critical in scenarios with scarce fault data.

Roundtrip probability estimation: A generative approach that maps between observed data and latent variables to estimate true sample distributions and guide synthetic data creation.

References

1. Tomek Link and SMOTE Approaches for Machine Fault Classification with an Imbalanced Dataset. Sensors (2022).
2. An Oversampling Method of Unbalanced Data for Mechanical Fault Diagnosis Based on MeanRadius-SMOTE. Sensors (2022).
3. Gradient-Oriented Prioritization in Meta-Learning for Enhanced Few-Shot Fault Diagnosis in Industrial Systems. Applied Sciences (2023).
4. A roundtrip probability estimation method for mechanical equipment fault detection under imbalanced samples. Measurement and Control (2024).