An Intermittent Fault Data Generation Method Based on LSTM and GAN
Junyou Shi, Yufei Ding, Zhenyang Lv · 2021 Global Reliability and Prognostics and Health Management (PHM-Nanjing) · 2021
Intermittent fault has the characteristic of degradation, and may eventually evolve into permanent failure. In some areas, the data of intermittent fault collected through actual operation cannot meet the large requirement for fault diagnosis and degradation assessment. This paper proposes a novel GAN model based on LSTM. Combined with GAN’s adversarial idea and LSTM’s ability to process time series. The generative model is able to automatically generate intermittent failure data under the condition of very few samples, so as to achieve the enrichment and supplement of intermittent failure data.