【摘要】 在液压系统智能诊断中，样本不平衡将使得智能诊断模型容易学习到样本充裕的液压系统健康状态的诊断规则，忽略样本匮乏的健康状态诊断规则，致使模型诊断精度下降。针对上述问题，提出基于自定步调集成的液压系统智能诊断方法。该方法首先将多数类样本集根据不同的硬度等级分为k个容器；然后通过欠采样平衡每个容器对分类硬度的贡献，使重采样后每个容器中的样本硬度总和一致；在训练过程中不断更新自步因子，用来降低样本数量过多的容器采样权重，经过n次迭代形成最终的集成诊断模型。将提出方法应用于液压系统不同故障数据集进行智能诊断，结果表明，该方法能够提高样本不平衡情况下液压系统故障识别的准确率，且诊断结果优于传统的智能故障诊断方法。更多还原

【Abstract】 In the intelligent diagnosis of hydraulic system, sample imbalance will make the intelligent diagnosis model easy to learn the diagnosis rules of hydraulic system health state with abundant samples, and ignore the health state diagnosis rules of short of samples, resulting in the decrease of model diagnosis accuracy. An intelligent diagnosis method of hydraulic system based on self-paced ensemble is proposed. The method first divides the majority class sample set into k bins according to different hardness levels, and then balances the contribution of each container to classification hardness through under-sampling, so that the total hardness of the samples in each container after resampling is consistent. In the training process, self-step factor is continuously updated to reduce the sampling weight of container with too many samples, and the final ensemble diagnostic model is formed after n iterations. The proposed method is applied to different fault data sets of hydraulic system for intelligent diagnosis. The results show that the proposed method can improve the accuracy of fault identification of hydraulic system in case of unbalanced samples, and the diagnosis results are better than traditional intelligent fault diagnosis methods.更多还原