基于实时可靠度评估的航空产品序贯检测模型

Sequential Inspection Model for Aviation Product Based on Real-timeReliability Evaluation

针对航空产品相关部件性能退化不确定性强的特点,为了较为科学合理地确定产品检测维修的时刻,提出一类基于实时可靠度评估的序贯检测模型。采用Wiener过程描述产品的性能退化规律,然后基于随机过程首达时理论给出产品实时可靠度的解析表达式,并综合利用强跟踪滤波算法、最优平滑算法和期望最大化算法对模型未知参数进行自适应估计。当获得新的退化数据信息时,根据贝叶斯规则重新迭代对模型参数进行实时更新。在更新过程中,融合同类型产品的历史信息选取初始化参数,实现了更新算法的快速收敛。接着根据对产品的实时可靠度要求确定出了优化的序贯检测间隔期。通过某航空铝合金材料的疲劳裂纹增长的实例验证了模型与算法的有效性。研究结果表明:实时可靠度评估模型能够最大限度利用运行期的实时数据,有效提高不确定条件下产品可靠度估计的准确性,从而保证相应序贯检测策略的效率和实用性。

Due to the strong uncertainty of degradation of relevant aviation components, a sequential inspection model based onreal-time reliability evaluation is proposed to determine the interval of inspection and maintenance more scientifically. The Wienerprocess is used to describe the performance degradation of products. The analytical expression of real-time reliability distribution isobtained by use of the first-hitting time theory. And an adaptive method is proposed to evaluate the unknown parameters by using thestrong tracking filtering algorithm, the optimal smoothing algorithm(RTS) and the expectation maximization algorithm(EM). Oncethe new degradation information is available, the parameters should be updated with Bayesian equation in real time. Moreover, thehistorical information of the same type product is integrated in the selection of the initial parameters to achieve the fast convergenceof the updating algorithm. Thirdly, according to the real-time reliability requirements of the product, the optimized sequentialinspection interval is determined. Finally, an example of fatigue crack growth in an aerospace aluminum alloy is given to illustrate thevalidity of the proposed model and the algorithm. The results have proved that the real-time data of aviation products during operationcan be fully utilized by the model based on real-time reliability evaluation, and the accuracy of product reliability estimation can beimproved effectively under uncertain conditions, so that the efficiency and practicability of the sequential inspection strategy areensured.