摘要
针对轴向柱塞泵故障诊断中因故障样本稀缺及数据分布不均衡导致的模型泛化能力不足的问题,提出一种基于Wasserstein生成对抗网络(WGAN-GP)进行数据增强的方法,以解决轴向柱塞泵故障数据提取难的问题。通过AMESim软件模拟不同轴向柱塞泵故障状态,并采集相应的流量信号和压力信号。利用高斯噪声和WGAN-GP对采集的数据进行增强,以提高样本数据的数量和质量。提取时域、频域和小波包能量特征作为初始特征参数,并将其输入长短期记忆网络(LSTM)进行故障模式识别和分类。结果表明:该方法在故障样本较少的情况下,能有效提高故障诊断的准确率和模型的泛化能力;对流量信号进行故障诊断时,数据增强前的最高识别准确率为86.84%,应用WGAN-GP算法对训练集数据进行数据增强后的识别准确率最高达98.72%;对压力信号进行故障诊断时,数据增强前的识别最高准确率为88.89%,应用WGAN-GP算法对训练集数据进行数据增强后的识别准确率最高达95.65%。数据增强算法对于提高故障诊断模型的识别准确率具有重大意义。
In order to solve the problem of insufficient model generalization ability caused by the scarcity of fault samples and uneven data distribution in the fault diagnosis of axial piston pump,a method for data augmentation based on the Wasserstein generative adversarial network with gradient penalty(WGAN-GP)was proposed,addressing the challenge of insufficient fault data extraction.The fault states of different axial piston pumps were simulated through AMESim software,and the corresponding flow and pressure signal were collected.The collected data were enhanced by using Gaussian noise and WGAN-GP to improve the quantity and quality of sample data.The time-domain,frequency-domain and wavelet packet energy features were extracted as the initial feature parameters,and they were input into the long short-term memory network(LSTM)for fault mode identification and classification.The results show that the proposed method can effectively improve the accuracy of fault diagnosis and the generalization ability of the model when there are few fault samples.The highest recognition accuracy before the enhancement of the flow signal is 86.84%,and the highest recognition accuracy of the WGAN-GP algorithm is 98.72% after the data enhancement of the training set data.The highest recognition accuracy before the enhancement of the pressure signal data is 88.89%,and the highest recognition accuracy of the WGAN-GP algorithm is up to 95.65%after the data enhancement of the training set data.The data-based augmentation algorithm is of great significance to improve the recognition accuracy of the fault diagnosis model.
作者
钟金豹
孟祥一
范浩熙
王永鹏
冯相龙
张剑
ZHONG Jinbao;MENG Xiangyi;FAN Haoxi;WANG Yongpeng;FENG Xianglong;ZHANG Jian(School of Mechanical Engineering,Inner Mongolia University of Science and Technology,Baotou Inner Mongolia 014000,China)
出处
《机床与液压》
北大核心
2025年第13期153-159,共7页
Machine Tool & Hydraulics
关键词
轴向柱塞泵
故障诊断
特征提取
数据增强
axial piston pumps
fault diagnosis
feature extraction
data augmentation
