关键词: 机理与数据双驱, 飞机系统, 可靠性评估, 生成对抗代理建模, 液压系统低压, 起落架刹车温度多失效

Abstract:

To effectively perform aircraft system reliability evaluation， the Mechanism and Data dual-drive Reliability Monitoring （MDRM） concept has been proposed. In the MDRM concept， the fault logic diagram is constructed from the perspective of forward mechanism based on the Functional Hazard Analysis （FHA）， Failure Mode and Effects Analysis （FMEA）， and Fault Tree Analysis （FTA）. By integrating optional data， a Bayesian network model is built to select important influencing parameters. The generative adversarial theory is introduced into the surrogate model， and a generative adversarial surrogate modeling strategy is presented to establish a correlation model between the influencing parameters and research object， thus enabling the reliability evaluation. The Generative Adversarial Regression Network （GARN） method is proposed for aircraft systems reliability evaluation by integrating the MDRM concept with neural network models and compact support region thought. In addition， the mathematical cases are adopted to demonstrate the modeling performance of proposed GARN， and the engineering applicability of developed method are verified through the No.1 hydraulic system low-pressure and landing gear brake temperature multi-failures of a domestic civil aircraft. The comparison of several methods shows that GARN holds outstanding modeling and simulation performance advantages， and the proposed concept and method can provide strong theoretical and technical support for aircraft system reliability assessment.

Key words: mechanism and data dual-drive, aircraft system, reliability evaluation, generative adversarial surrogate modeling, hydraulic system low-pressure, landing gear brake temperature multi-failures