针对传统可靠性分析方法必须依赖大样本统计数据、利用概率统计求解设备可靠性的不足,提出两种利用运行状态信息实现小样本条件下设备运行可靠性评估的方法:基于归一化小波信息熵的可靠性评估和基于损伤定量识别的可靠性评估。基于归一化小波信息熵的可靠性评估对设备运行过程中的振动信号采用第二代小波包进行分解与重构,获得多个分解频带信号,计算分解频带信号的相对能量和归一化信息熵,根据归一化信息熵获取反映设备运行状态可靠性的重要指标——可靠度;基于损伤定量识别的可靠性评估定义了新的运行可靠性评价指标——隶属可靠度,在故障定量诊断的基础上,建立了损伤程度与运行可靠性评价指标之间的联系。在制氧压缩机运行可靠性评估和机车轮对轴承运行可靠性评估的成功应用,表明所提出的方法合理、有效,为机械设备实现缺乏大样本、非概率统计模型的可靠性评估提供了新理论与新技术。
The traditional reliability analysis has such shortcomings that it relies on probability statistics with the large sample statistical data. This study proposes two operation reliability assessment methods which use running condition monitoring information to realize the reliability evaluation under small sample. They are there- liability evaluation methods based on the normalized wavelet information entropy and the damage quantitative identification respectively. Vibration signals of mechanical equipment are decomposed and reconstructed by means of second generation wavelet package to acquire decomposed signals in sub-frequency bands, so that full condition information of running equipment can be adequately used. We take relative energy in each sub-frequency band to calculate normalized information entropy. The reliability degree, an important reliability index, is trans- formed by using the normalized wavelet information entropy to assess operation reliability for running equip- ment. A new operation reliability assessment index called membership reliability degree is defined and an operation reliability assessment model is built based on quantitative damage diagnosis. Successful applications have been achieved to assess operation reliability of an oxy-generator compressor and the wheel bearings in electric locomotives, which demonstrated that the proposed approaches are reasonable and effective. The paper provides new approaches without large sample size, which are independent of probability to operation reliability assessment for large machinery.