基于深度学习的飞行器故障不确定性评估与预测研究

项目名称： 基于深度学习的飞行器故障不确定性评估与预测研究

项目编号： No.51475368

项目类型： 面上项目

立项/批准年度： 2015

项目学科： 机械、仪表工业

项目作者： 姜洪开

作者单位： 西北工业大学

项目金额： 83万元

中文摘要： 飞行器关键部件状态变化频繁和工作环境复杂多变，使得飞行器故障具有不确定性，而关键部件故障不确定性往往会导致灾难性后果，飞行器故障预测与健康管理一直是研究热点。本项目针对飞行器关键部件故障不确定性评估与预测这一难点与重点问题，探索飞行器关键部件故障不确定性机理，建立飞行器关键部件故障不确定性模型，揭示飞行器关键部件故障评估和预测中的不确定性因素；建立基于多源飞行参数信息的飞行器关键部件故障不确定性信息度量模型，构建不确定性故障特征量综合评估体系；研究基于不确定性故障特征量的特征表示和深度学习模型建立方法，构造基于学习深度方法的飞行器关键部件故障不确定性评估和预测算法。对叶片、机翼、轴、轴承等飞行器关键部件开展试验验证与应用研究。本项目研究可望在故障不确定性分析理论、飞行器关键部件故障不确定性评估与预测方法方面有所创新和突破，为提升飞行器的安全性和可靠性提供新的理论与方法。

中文关键词： 飞行器；故障不确定性；评估与预测；深度学习；故障预测与健康管理

英文摘要： The operational conditions of aerial vehicle key parts change frequently and the load environment is complicated and variable, which lead to aerial vehicle parts fault uncertainty. Aerial vehicle key part fault usually is disaster. Aerial vehicle fault prognosis and health management always is the research focus. This project aims at the key problem of aerial vehicle key part fault evaluation and prognosis. The dynamic evolution mechanism of aerial vehicle fault uncertainty is studied. The uncertainty simulation model of aerial vehicle key part is constructed to reveal the uncertainty factors for aerial vehicle fault evaluation and prognosis. Aerial vehicle fault uncertainty measurement model based on multi-flight parameters is set up, and the integration system of uncertainty fault feature indicator vector is constructed. The deep learning model is selected based on uncertainty fault feature indicator vector, and the deep learning dynamic evaluation and prognosis algorithms for aerial vehicle fault uncertainty are constructed. The experimental verification and application for blade, aircraft wing, shaft, bearing and so on are carried out. The novel results and contribution in fault uncertainty analysis theory and aerial vehicle fault uncertainty dynamic evaluation and prognosis will be obtained. This project research will provide new theory and technology in enhancing aerial vehicle safety and reliability.

英文关键词： Aerial Vehicle;Fault Uncertainty;Evaluation and Prognsosis;Deep Learning;Fault Prognosis and Health Management