摘要详情

ID / 提交时间

77 / 2025-04-01 16:35:47

标题

Rolling Bearing Fault Diagnosis Based on OCSSA-VMD-CNN-GRU

关键字

rolling bearing, fault diagnosis, variational modal decomposition, convolutional neural network, gated recurrent unit

主题及专题

T01 模型驱动的故障诊断

状态

全文待审

作者

邦成张 / 长春工业大学;长春工程学院

博文胡 / 长春工程学院

波李 / 长春工程学院

婧媛宋 / 长春工程学院

云高殷 / 长春工程学院

智高 / 长春工业大学

摘要

Abstract—To address the challenge of identifying early rolling bearing faults under complex operating conditions, this paper proposes a hybrid VMD-CNN-GRU model optimized using the Osprey-Cauchy-Sparrow search algorithm (OCSSA). First, we design the OCSSA algorithm by integrating the global optimization capability of the Ospreys Search with the local perturbation mechanism of Cauchy mutation, enabling adaptive optimization of both the penalty factor and mode number in variational mode decomposition (VMD). This effectively mitigates modal aliasing issues. Subsequently, an end-to-end deep diagnostic framework is constructed to extract the time-frequency fusion features of the decomposed signals using CNNs and introduce gated recurrent units (GRUs) to model long-term dependencies in the sequences, thereby improving the recognition of weak fault features. Experimental validation on the Case Western Reserve University (CWRU) bearing dataset demonstrates the superiority of the proposed model over benchmark methods (e.g., VMD-CNN-LSTM, VMD-CNN-GRU,and OCSSA-VMD-CNN-BiLSTM),achieving 99.67% fault recognition accuracy and 3.92-second diagnostic latency, which highlights its strong generalization capability and engineering applicability.