中文摘要：

针对表征齿轮故障信息的特征难提取与极限学习机无法处理随时间变化的信息流，致使齿轮故障分类模型精度差、泛化能力弱的问题，提出一种基于在线贯序极限学习机的齿轮故障诊断方法。该方法首先将齿轮振动信号进行相空间重构，并对重构矩阵进行奇异值分解得到奇异值特征向量；其次，建立在线贯序极限学习机的齿轮故障分类模型，并将奇异值特征向量作为模型输入进行齿轮不同故障状态的辨识。实验结果表明：与基于BP、SVM与ELM的故障分类方法相比，基于基于在线贯序极限学习机的齿轮故障诊断方法具有更快的学习速度、更高的分类精度与更强的泛化能力。

英文摘要：

In order to solve the problems that gear fault classification model has poor accuracy, weak generalization ability causing by the fanlt features of gear is difficult to extract and extreme learning machine is unable to process information flow that changes over time, this paper puts forward a method of gear fault diagnosis based on online sequential extreme learning machine. First, this method reconstructs the gear vibration signals in phase space, and decomposes the reconstruction matrix into singular value feature vectors. Secondly, the online sequential extreme learning machine is used to set up geac fault classification model and singular value feature vectors is imported into the model to identificate different gear fault states. The experimental results show that compared with the method of BP, SVM and ELM, the method of gear fault diagnosis based on online sequential extreme learning machine has faster learning speed, higher classification accuracy and stronger generalization ability.