基于BO-DKELM的滚动轴承故障诊断

滚动轴承作为旋转机械中的必需元件，其任何故障都可能导致机器乃至整个系统发生故障，从而导致巨大的经济损失和时间的浪费，因此必须要及时准确地诊断滚动轴承故障。针对传统极限学习机中模型参数对滚动轴承故障诊断精度影响较大的问题，提出了一种基于贝叶斯优化的深度核极限学习机的滚动轴承故障诊断方法。首先，将自动编码器与核极限学习机相结合，构建了深度核极限学习机(Deep kernel extreme learning machine, DKELM)模型。其次，利用贝叶斯优化(Bayesian optimization, BO)算法对DKELM中的超参数进行寻优，使得训练数据集和验证数据集在DKELM模型中的分类错误率之和最低。然后，将测试数据集输入到训练好的BO-DKELM中进行故障诊断。最后，采用凯斯西储大学轴承故障数据集对所提方法进行验证，最终故障诊断精度为99.6%，与深度置信网络和卷积神经网络等传统智能算法进行对比，所提方法具有更高的故障诊断精度。

Rolling Bearing Fault Diagnosis Based on BO-DKELM

As a necessary component in rotating machinery, any failure of rolling bearings may lead to the failure of the machine or even the whole system, which leads to huge economic loss and time wastage. Therefore,it is necessary to diagnose the rolling bearing fault promptly and accurately. In response to the problem that the model parameters in the traditional extreme learning machine have a large influence on the fault diagnosis accuracy of rolling bearings, a rolling bearing fault diagnosis method based on a deep kernel extreme learning machine with Bayesian optimization is proposed. Firstly, the deep kernel extreme learning machine (DKELM) model is constructed by combining the auto encoder (AE) with the kernel extreme learning machine (KELM). Secondly, a Bayesian optimization algorithm is used to find the optimal hyperparameters in the DKELM, such that the sum of the classification error rates of the training and validation datasets in the DKELM model is minimized. The test dataset was then fed into the trained BO-DKELM for fault diagnosis. Finally, the proposed method was validated using the Case Western Reserve University bearing fault dataset, and the final fault diagnosis accuracy is 99.6%, comparing with traditional intelligent algorithms such as deep belief networks and convolutional neural networks, the proposed method has higher fault diagnosis accuracy.