WBLG: A method for bearing fault diagnosis based on the fusion of spatial and temporal features

Abstract Bearing fault diagnosis is a key strategy to ensure the stability of mechanical systems, optimize maintenance plans and improve operational reliability. Vibration signals are complex time series with unique properties. Most of the current methods only consider the spatial characteristics of the signal, but do not take into account its temporal characteristics. In fact, vibration signals contain both spatial and temporal information, offering not only rich temporal dynamic details but also spatial structural insights that reflect fault characteristics. Therefore, in oder to fully exploit the spatial and temporal information in raw one-dimensional vibration signals, this study proposes a progressive spatio-temporal feature fusion framework, termed WBLG. In this framework, Wide kernel deep convolutional neural network(WDCNN) is used as a front-end module to extract local impact-related spatial features directly from raw vibration signals. An ECA-based channel-adaptive module is then embedded after the first wide convolutional layer to enhance fault-sensitive channels and suppress redundant responses. Subsequently, a cascaded BiLSTM GRU module is employed to capture and refine temporal dependencies from the enhanced feature sequence. This design preserves the original temporal structure and transient impact characteristics of vibration signals while progressively integrating spatial, channel-wise, and temporal fault information. To verify the proposed method, ablation and comparative experiments were conducted on two publicly available bearing datasets: the Case Western Reserve University dataset and the Guangdong University of Petrochemical Technology dataset. The proposed WBLG network attains average fault diagnosis accuracies of 99.7% and 96.5% on the two datasets respectively, with all reported results representing the mean across five independent trials. Compared with the existing models, the maximum improvement rates are 3.73% and 12.4% respectively. This results demonstrate the superior classification performance and generalization applicable to bearing fault diagnosis.