Fault Detection and Diagnosis with Imbalanced and Noisy Data: A Hybrid Framework for Rotating Machinery

Fault diagnosis plays an essential role in reducing the maintenance costs of rotating machinery manufacturing systems. In many real applications fault detection and diagnosis, data tend to be imbalanced, meaning that number samples for some classes is much less than normal samples. At same time, industrial condition, accelerometers encounter high levels disruptive signals collected turn out heavily noisy. As a consequence, traditional Detection Diagnosis (FDD) frameworks get poor classification performances when dealing with real-world circumstances. Three main solutions have been proposed literature cope this problem: (1) implementation generative algorithms increase amount under-represented input samples, (2) employment classifier being powerful learn from imbalanced noisy data, (3) development efficient pre-processing including feature extraction augmentation. This paper proposes hybrid framework which uses three aforementioned components achieve effective signal-based FDD system conditions. Specifically, it first extracts features, using Fourier wavelet transforms make full use signals. Then, employs Wasserstein Generative Adversarial Networks (WGAN) generate synthetic populate rare class enhance training set. Moreover, higher performance novel combination Convolutional Long Short-term Memory (CLSTM) Weighted Extreme Learning Machine (WELM) proposed. To verify effectiveness developed framework, different datasets settings on imbalance severities noise degrees were used. The comparative results demonstrate scenarios GAN-CLSTM-ELM outperforms other state-of-the-art frameworks.