基于生成对抗网络数据增强的舰炮可靠性分析OA北大核心CSTPCD
Reliability Analysis of Naval Guns Based on GAN Data Augmentation
舰炮故障数据不足会导致其可靠性分析变得极其困难.为解决故障数据不足这一问题,采用生成式对抗网络(generative adversarial network,GAN)对故障样本进行数据增强.建立GAN数据增强深度神经网络的舰炮可靠性分析模型,并与原始数据进行可靠性分析获得的指标进行了对比.结果表明,利用GAN数据增强后的扩充样本得到的指数分布拟合精度及威布尔分布拟合精度分别提高了 5.40%和 11.90%,相较于原始数据有了显著提升.为实现舰炮故障数据可靠性分析提供了方法和思路.
The lack of fault data of naval guns makes it extremely difficult to analyze the reliability of products.In order to solve the problems of lacking fault data,the generative adversarial network(GAN)is used to augment the fault data.The reliability analysis model of naval guns with GAN data augmen-tation deep neural network is established.The comparison is carried out with indicators obtained from the reliability analysis of original data.The results show that the exponent distribution fitting accuracy improves by 5.40%and Weibull distribution fitting accuracy improves by 11.90%respectively by the expanded samples with GAN data augmentation.Compared with the original data,there is notable im-provement.The methods and ideas are provided for the reliability analysis of fault data of naval guns.
聂磊;杨浩明;尹业寒;董正琼;周向东
湖北工业大学机械工程学院,武汉 430068||湖北省现代制造质量工程重点实验室,武汉 430068湖北工业大学机械工程学院,武汉 430068
舰炮故障小样本GAN数据增强可靠性
naval gun failuresmall sampleGANdata augmentationreliability
《火力与指挥控制》 2024 (004)
38-43,50 / 7
国家自然科学基金(51975191);襄阳湖北工业大学产业研究院基金资助项目(XYYJ2022B01)
评论