基于SMOTE-IPF和SDAE的行星齿轮箱故障诊断方法研究

Research on Fault Diagnosis Method of Planetary Gearbox Based on SMOTE-IPF and SDAE

针对现有故障诊断模型在故障样本缺乏时诊断率低的问题,提出一种基于迭代过滤合成少数类过采样方法(SMOTE-IPF)和堆叠去噪自动编码器(SDAE)的故障诊断模型。该方法利用SDAE对样本进行特征提取,使用SMOTE-IPF在合成新样本的同时利用多个决策树对新样本进行投票过滤,使数据集达到平衡,最后使用分类器进行故障分类。通过行星齿轮实验平台进行实验,验证了所提方法在故障样本极度缺乏下故障诊断的有效性。

To solve the problem of low diagnosis rate of existing fault diagnosis models when in lack of fault samples,this paper proposes a fault diagnosis model based on iterative filtering synthetic minority class oversampling method(SMOTE-IPF)and Stacked Denoising AutoEncoder(SDAE).This method uses SDAE to extract features from samples,applies SMOTE-IPF to synthesize new samples and uses multiple decision trees to vote and filter new samples to balance the dataset,and adopts a classifier to classify faults.Experiments are carried out on the planetary gear experimental platform to verify the effectiveness of the proposed method in fault diagnosis in the extreme lack of fault samples.