A Novel Graph Integrated Selective State Space Fault Network for Intelligent Gearbox Fault Diagnosis Using Vibration Signal Analysis

ABSTRACT

Gearboxes play a vital role in numerous industrial machines that include wind turbines, automotive transmissions, manufacturing systems, etc. Heavy mechanical and continuous operation result in different types of faults such as cracks, tooth breakage, and gear wear. Early detection of these faults is essential to reduce maintenance costs, machine failures, and production downtime. Hence, reliable gearbox fault diagnosis has become a significant research area in intelligent condition monitoring. To address such shortcomings, this paper proposes a novel graph‐integrated selective state space fault network (GSS‐FaultNet) for gearbox fault diagnosis using vibration signal analysis. The proposed model initially performs signal preprocessing to enhance data quality. The comprehensive fault features generate time domain and frequency domain representations. A graph construction module is introduced to model the structural relationship among gearbox components as well as sensor locations. Simultaneously, a selective state space module effectively captures long‐range temporal dependency from various sequential vibration signals. Finally, the extracted temporal features are fused via a feature integration layer, followed by a fully connected layer for classifying faults. Thus, the hybrid architecture enables the model to capture the temporal dynamics and structural interactions of gearbox behavior. Experimental validation demonstrates that the proposed GSS‐FaultNet framework attains enhanced fault classification accuracy. The results indicate that the proposed model enhances the capability of gearbox health monitoring systems. The proposed model demonstrates an effective capability to learn both temporal and structural features, achieving the highest accuracy of 97.7%.