生成对抗网络的多联机四通阀故障数据扩增策略及故障诊断

Fault Data Amplification Strategy and Fault Diagnosis for Multi-line Four-way Valve by Generative Adversarial Nets

故障数据不足已经成为多联机故障诊断技术发展的主要障碍。本文采用生成对抗网络,对多联机四通阀故障数据进行了扩增,有效避免了数据驱动模型训练集数据不均衡问题,分析了引入生成对抗网络扩增的数据对数据驱动模型诊断准确率的影响。结果表明,引入生成对抗网络生成的多联机四通阀故障数据后,减缓了故障数据量和正常数据量的不平衡程度,测试集整体诊断准确率由92.29%提升至97.00%,几何诊断准确率由21.12%提升至97.13%。

Insufficient fault data have become a major obstacle to the development of multi-online fault diagnosis technology. In this paper, a generative confrontation network is used to reasonably and effectively amplify the fault data of the multi-line four-way valve, which effectively avoids the problem of data imbalance in the training set of the data-driven model, and analyzes the impact of introduction of the amplified data of the generative confrontation network on the diagnostic accuracy of the data-driven model. The results show that the introduction of the multiline four-way valve fault data generated by the generative confrontation network reduces the imbalance between the amount of fault data and the amount of normal data. The overall diagnostic accuracy rate of the test set is increased from 92.29% to 97.00%, and the geometric diagnostic accuracy rate is increased from 21.12% to 97.13%.