基于改进深度森林的电动执行器故障诊断方法

Fault Diagnosis Method of Electric Actuator Based on Improved Deep Forest

针对现有深度学习方法在电动执行器故障诊断中训练数据需求量大、调参困难问题,提出一种基于改进深度森林的电动执行器故障诊断方法。首先,利用时频域信号特征提取方法处理样本数据,并采用多粒度扫描提取多级特征;然后,将D-S证据理论与级联森林结合,克服在扩展级联森林过程中遇到的特征冗余问题;最后,采用改进深度森林方法构建了故障诊断模型,通过超参数试验和对比试验验证所提方法。结果表明:所提方法不仅能在样本数据较少时有效诊断故障类别,而且在不同工况下,表现出较好的泛化能力。

Aiming at the problems of large training data demand and difficulty in tuning parameters in the fault diagnosis of electric actuators with existing deep learning methods,a fault diagnosis method of electric actuators based on improved deep forest was proposed.Firstly,the time-domain and frequency-domain features extraction method was used to process the sample data,and the multi-granularity scan was used to extract multi-level features.Then,the D-S evidence theory is combined with the cascade forest to overcome the feature redundancy problems encountered in the process of expanding the cascade forest.Finally,an improved deep forest method is used to construct the fault diagnosis model,and the proposed method is verified through hyperparameter experiments and comparative experiments.The results illustrate that the proposed method can not only effectively diagnose the fault category with a small amount of sample data,but also show good generalization ability under different working conditions.