【摘要】 针对噪声环境下滚动轴承故障诊断困难且诊断模型参数量过大的问题，提出了一种联合支持向量机的轻量化卷积神经网络(light-weight convolutional neural network-support vector machine, LWCNN-SVM)模型用于滚动轴承的故障诊断。首先，该模型采用通道分离的多特征输出结构与全局平均池化层来降低模型参数，并直接从噪声信号中提取特征。随后，采用SVM对低维特征进行分类，从而实现滚动轴承的故障诊断。在不同噪声环境和样本量条件下，与传统的方法对比，验证了所提方法具有良好的抗噪性能和较低的计算成本。更多还原

【Abstract】 To address the challenges of rolling bearing fault diagnosis in noisy environments and the high computational cost associated with large-scale diagnostic models, this paper proposes a novel lightweight fault diagnosis method combining a Convolutional Neural Network with a Support Vector Machine(LWCNN-SVM). The model leverages a multi-feature output structure with channel separation and a global average pooling layer to significantly reduce the number of parameters. The LWCNN is designed to extract meaningful features directly from raw noisy signals, while the SVM classifier is employed to perform fault classification using the resulting low-dimensional features. Comparative experiments conducted under varying noise levels and sample sizes confirm that the proposed method demonstrates strong noise robustness and low computational overhead, making it well-suited for practical engineering applications.更多还原