【摘要】 针对实际工况中轴承故障性能退化难以准确评估的问题，提出了一种基于连续隐马尔可夫模型(CHMM)的性能指标(PI)评价框架。该框架首先通过PI评价指标对轴承退化信息进行定量分析，并据此将故障劣化过程划分为3个特征明显的阶段。随后，采用双谱分析和1.5维谱分析方法对故障劣化程度进行识别，发现其随轴承劣化加深呈现出显著的趋势性和一致性特征。在模型应用方面，针对CHMM在训练过程中存在的初始模型参数选取困难和算法下溢问题，提出了等概率观测值矩阵初始化方法和比例因子修正算法。通过某大学轴承全生命周期实验数据的验证，结果表明所提出的方法不仅能够准确评估轴承劣化程度，还能有效捕捉状态退化的突变点，清晰揭示轴承退化状态的发展趋势，为轴承性能退化评估提供了新的技术手段。更多还原

【Abstract】 To address the challenge of accurately evaluating bearing fault degradation under actual working conditions, this study proposes a performance indicator(PI) evaluation framework based on the continuous hidden markov model(CHMM). The framework first performs a quantitative analysis of bearing degradation information using PI evaluation metrics, dividing the degradation process into three distinct phases. Subsequently, bispectral analysis and 1.5D spectral analysis methods are employed to assess the degree of degradation, revealing significant trends and consistency as the bearing deterioration progresses. In terms of model application, this study tackles issues related to the selection of initial model parameters and algorithm underflow during CHMM training by introducing an equal-probability observation matrix initialization method and a proportional factor correction algorithm. Validation using full lifecycle bearing data from a certain university demonstrates that the proposed methods not only accurately evaluate the extent of bearing degradation but also effectively capture abrupt changes in the degradation state. Furthermore, the approach clearly reveals the development trend of bearing degradation, providing a novel technical tool for evaluating bearing performance degradation.更多还原