Po-Jui Chiang

Robust Unsupervised Acoustic Anomaly Detection for Turbo Molecular Pumps Using ResNet–Convolutional Block Attention Module and Structural Similarity Loss

An unsupervised acoustic anomaly detection and localization system to address real-time predictive maintenance in noisy industrial settings and achieves an outstanding AUC of 0.9535 and a fault recall of 99.06% under a validation-calibrated threshold, significantly outperforming standard U-Net architectures.