Hypergraph Contrastive Sensor Fusion for Multimodal Fault Diagnosis in Induction Motors

TL;DR

Induction motor fault diagnosis faces challenges from multimodal sensor signals, noise, and cross-domain shifts. The paper introduces MM-HCAN, a unified framework that uses dual-path feature extraction from raw signals and STFT spectrograms, hypergraph-based intra- and cross-modality fusion, contrastive learning, and multi-head attention to achieve robust, generalisable fault classification across bearing, stator, and rotor faults. Key contributions include a hypergraph neural network with Laplacian-based embedding updates, a triplet loss within the hypergraph topology, and a fused representation for final classification, demonstrated on three benchmarks with state-of-the-art accuracy and strong noise and cross-domain resilience. The approach offers a scalable path toward real-time predictive maintenance in industrial settings, enabling consolidated multi-fault diagnosis and extended asset longevity.

Abstract

Reliable induction motor (IM) fault diagnosis is vital for industrial safety and operational continuity, mitigating costly unplanned downtime. Conventional approaches often struggle to capture complex multimodal signal relationships, are constrained to unimodal data or single fault types, and exhibit performance degradation under noisy or cross-domain conditions. This paper proposes the Multimodal Hypergraph Contrastive Attention Network (MM-HCAN), a unified framework for robust fault diagnosis. To the best of our knowledge, MM-HCAN is the first to integrate contrastive learning within a hypergraph topology specifically designed for multimodal sensor fusion, enabling the joint modelling of intra- and inter-modal dependencies and enhancing generalisation beyond Euclidean embedding spaces. The model facilitates simultaneous diagnosis of bearing, stator, and rotor faults, addressing the engineering need for consolidated di- agnostic capabilities. Evaluated on three real-world benchmarks, MM-HCAN achieves up to 99.82% accuracy with strong cross-domain generalisation and resilience to noise, demonstrating its suitability for real-world deployment. An ablation study validates the contribution of each component. MM-HCAN provides a scalable and robust solution for comprehensive multi-fault diagnosis, supporting predictive maintenance and extended asset longevity in industrial environments.

Figures (7)

• Figure 1: MM-HCAN architecture: temporal features from raw signals (1D CNN-LSTM) and spectral features from STFT images (ResNet-18) are fused via hypergraph contrastive learning and multi-head attention for classification.
• Figure 2: Hypergraph construction: intra- and cross-modality hyperedges are formed via KNN to model high-order relationships among feature vectors across modalities.
• Figure 3: STFT analysis of various motor component conditions: (a-d) Bearing health states, (e-h) Stator fault conditions and rotor health, (i-l) BRB fault scenarios. All plots share identical time-frequency scales.
• Figure 4: Confusion Matrix of Bearing Vibration Dataset
• Figure 5: Confusion matrices: (a) Stator Current Analysis (b) Stator Vibration evaluation using MM-HCAN model.