针对健康指标构建方法依赖专家经验的问题,本文结合MSR-CNED-SE提出了一种基于无监督数据集的健康指标构建方法,并利用深度学习模型实现剩余寿命预测。通过在轴承全寿命数据集上验证方法的可靠性。此外,还通过研究不同相似度与深度学习网络对健康指标的影响。结果表明,基于平方欧式距离作为相似度度量构建的指标更容易找到退化的起始点,而Bi-LSTM网络在不同预测场景下表现出更好的稳定性和可靠性。

In response to the problem of relying on expert knowledge in health indicator construction methods, this paper proposes a method for constructing health indices based on unsupervised deep learning techniques, integrated with MSR-CNED-SE. Furthermore, it utilizes deep learning models to predict remaining useful life. The reliability of the proposed method is validated on a comprehensive bearing life dataset. Additionally, the impact of different similarity measures and deep learning networks on the performance of health indices is investigated. The results indicate that health indices constructed using the Euclidean squared distance as a similarity measure are more effective in identifying the onset of degradation. Moreover, the Bi-LSTM network exhibits better stability and reliability across different prediction scenarios. This study provides a novel approach to health index construction that reduces the reliance on expert knowledge, and offers a feasible deep learning framework for remaining useful life prediction. These findings have significant theoretical and practical implications for achieving unsupervised health index construction and state-based maintenance decision-making.

TK83

西门子超超临界机组不稳定振动影响因素分析及对策研究