系统可靠寿命的不确定性分析及其高效方法

为了分析元器件失效率的不确定性对系统可靠性的影响,借鉴Borgonovo的矩独立灵敏度分析思想,在充分考虑了系统可靠寿命完整不确定性信息的情况下,提出了基于系统可靠寿命的矩独立重要性测度,用来分析不确定性条件下系统元器件失效率对其可靠寿命的平均影响。但由于系统可靠寿命函数是系统可靠度函数的反函数,一般无法解析表达而以隐函数的形式存在,致使该矩独立重要性测度难以高效准确求解。为了解决这一问题,文章提出了一种新的Kriging自适应代理模型的高效算法,该算法以Kriging代理模型预测值的变异系数作为自适应学习函数,通过自主增加新的试验样本,增强代理模型的预测准确性。阀门控制系统和民用飞机电液舵机系统两个算例分析表明,在保证计算精度的情况下,通过变异系数自适应学习函数,仅需添加少量系统可靠寿命试验样本,就能够构建用来充分近似系统可靠寿命函数的Kriging代理模型,解决了重要性测度的高效求解问题,从而验证了所提方法的合理性和算法的高效性。

Uncertainty analysis of systematic reliability life and its efficient solution

In order to study the influence of uncertain component failure rates to systematic reliability, a new systematic reliability life-based moment-independent importance measure was presented to analyze the components average impact of component failure rate under uncertainty to systematic reliability life. Inspired by the idea of the Borgonovo moment-independent sensitivity analysis, the proposed method fully takes the complete uncertainty information of systematic reliability life into consideration. Since the moment-independent importance measure was hardly solved accurately due to the implicit format of the inverse function of systematic reliability life function to systematic reliability function, therefore, a new method for Kriging adaptive surrogate model solving was proposed to improve the model prediction precision by adopting the response variation coefficient as the adaptive learning function and automatically increasing new samples. The two test cases of the valve control system and the civil aircraft electro-hydraulic actuator system results show that in the premise of computation accuracy, the Kriging model of systematic reliability life function can be fully approximated by adding small number of systematic reliability life test samples to the variation coefficient adaptive learning function. Hence, the new Kriging model successfully solves the importance measure problem, and the rationality of the proposed method and the high efficiency of the new algorithm are therefore verified.