新型大气数据传感系统的状态自确认方法研究

The distributed air data sensing technology, which is one of the most advanced flight parameters measuring techniques in the world, can be employed in the flight environments with high speed, high maneuverability or large angle of attack, and fine aircraft stealth design. As the source of acquiring flight parameters, the measurement value quality of the air data system will influence the control precision of aircraft control systems directly, and even threaten the flight safety. To enhance the reliability of measurement results, a novel air data sensing system is proposed in this project, in which the merits of the distributed air data sensing system have been fully combined. Further, the self-validating technology is used to implement the faults self-detection and self-diagnosis in real-world operation, the failure can be substituted by an optimal estimated value to implement the data self-recovery, and the important parameters about the working state can be provided to complete the status self-estimation. Based on the small and non-linear fault features which are extracted by using the sparse kernel principal component analysis, the rapid faults self-diagnosis method is studied by employing the multiclass relevance vector machine, in which multiple fault patterns can be classified simultaneously. Based on the redundant information of the distributed pressure points, the on-line data self-recovery method of multiple flight parameters is investigated by using the multi-variable relevance vector machine, in which it can buy valuable time to take the emergency actions. The different types of faults can bring the different influence on the accuracy of the validated measurement value, and then the on-line uncertainty estimation method is researched deeply based on the random fuzzy variable. The experimental platform is designed to collect data samples, and further to verify and improve the validity and real-time performance of the above self-validating algorithms. The prototype of the novel air data sensing system is developed, and this project can produce the state-of-the-art results on the online status self-validation, which will play a significant role in the development of the self-validating theory and the air data sensing system.

分布式大气数据传感技术是目前国际上最先进的飞行参数测量技术之一，可满足现代飞行器的高速飞行、高机动性及隐身需求。作为飞行参数信息获取的源头，其测量值的质量直接影响飞行控制的精度，甚至威胁飞行安全。为提高测量结果的可信度，本项目提出一种新型大气数据传感系统，它不但继承了分布式大气数据传感的技术优势，而且融合了自确认理论可实现故障自检测、自诊断，并利用最佳估计值代替故障数据实现数据自恢复，同时给出自身工作状态实现状态自评估。结合稀疏核主元分析提取的小样本、非线性故障特征，研究基于多模式分类相关向量机的快速故障自诊断方法；利用分布式测压点间的相关性，研究基于多变量相关向量机的飞行参数在线数据自恢复方法；结合不同的故障模式对测量值准确度的差异性影响，研究基于随机模糊变量的在线不确定度计算方法；搭建实验系统，进一步验证和完善各种状态自确认方法。本项目对自确认理论和大气数据传感系统的发展具有重要意义。

新型大气数据系统融合了分布式大气传感系统技术和自确认传感器技术的优点，它不但可给出原始测量信息，而且可对自身的工作性能和状态进行在线评估，实现故障自检测、自诊断，并利用最佳估计值代替故障数据实现容错处理，给后级飞行控制系统输出更加可靠的、丰富的信息，从而大大提高飞行器飞行安全。本项目旨在研究大气数据系统的状态自确认理论方法，解决大气数据系统的若干关键技术问题。基于专家经验知识，总结了大气数据系统中典型的故障模式，分析了各类故障的数据特征，研究了基于模糊概率Petri网的大气数据系统组件级及系统级故障传播分析方法；为解决大气数据系统的故障检测及故障源定位隔离问题，研究了基于小波核主元分析（WKPCA）和故障指示向量（FIV）的大气数据系统故障检测及故障源识别算法，该方法对典型故障的检测率优于90%，故障测压点均准确无漏被识别；为解决大气数据系统的故障自诊断问题，研究了基于集成经验模态分解（EEMD）的大气数据系统非线性故障特征提取算法，研究了基于多分类相关向量机（MCRVM）的大气数据系统故障诊断算法，该方法对典型故障的平均分类概率优于80%，且分类结果通过概率形式揭示了故障诊断结果的不确定性；针对大气数据系统在故障状态下的容错及状态评估问题，研究了基于多核相关向量机（MKRVM）的压力测量值数据恢复算法，单故障样本的恢复时间约3ms；研究了基于三点法的大气参数解算方法，研究了基于随机模糊向量（RFV）的大气数据系统不确定度在线估计算法，该方法给出了不同的置信水平及对应的置信区间，信息更丰富，结果更客观。从实际工程需求出发，扩展研究了大气数据系统贮存寿命加速试验评估及加速因子获取方法。为验证所研究算法的有效性，搭建了故障仿真试验平台，完成了压力传感器标定试验、大气系统正常测试试验、故障仿真试验及风洞试验，获取了算法验证的试验样本数据，同时也为新型大气数据系统的后续实用化奠定了基础。