摘要
基于相似性的剩余寿命预测方法是近年来兴起的一类部件剩余寿命预测方法,它可以提供合理的剩余寿命预测同时无需进行衰退过程建模(或仅需少量的相关工作)。针对当前基于相似性的剩余寿命预测方法仅利用失效的历史样本的问题,提出一个基于相似性的剩余寿命预测框架,同时利用失效与未失效(由于预防维护或终止使用等)的历史样本。在该框架中,提出两种估计未失效历史样本寿命,进而利用其衰退过程信息的方案(记为方案A和方案B)。一个系统的数值试验验证了在存在有限失效历史样本与存在大量失效历史样本的情况下,基于方案A的框架始终优于对应的传统方法。此外,试验结果揭示基于方案B的框架在失效历史样本有限的情况下并不有效,但随着失效历史样本的增多其表现迅速提升。数值试验的发现建议,在失效历史样本有限的情况下使用基于方案A的框架,而在可用大量失效历史样本时使用基于方案B的框架。
Similarity-based residual life prediction （SbRLP） approaches are emerging residual life （RL） prediction techniques that can provide proper RL predictions with few/no efforts on degradation modeling. Reported works on SbRLP approaches only utilize failed historical samples. In this paper, a framework of SbRLP approaches using both historical samples that failed and non-failed （due to preventive maintenance or suspension） is p framework, two solutions （namely Solution A and Solution B） are proposed to estimate the lifetimes of the non-failed historical samples, and to utilize their degradation histories. Using an extensive numerical experiment, it is demonstrated that the proposed framework based on Solution A is better than the corresponding classical SbRLP approach in the cases of limited failed historical samples and abundant failed historical samples. In the meantime, the investigation results revealthat the proposed framework based on Solution B is ineffective when failed historical samples are limited, but its performance improves fast with the increase of available failed historical samples. It is recommended to use the Solution A of the proposed framework when failed historical samples are limited and the Solution B when abundant historical samples are available.
出处
《电子产品可靠性与环境试验》
2012年第3期40-48,共9页
Electronic Product Reliability and Environmental Testing
