鲁棒的数据驱动的飞机故障诊断技术

项目名称： 鲁棒的数据驱动的飞机故障诊断技术

项目编号： No.61202078

项目类型： 青年科学基金项目

立项/批准年度： 2013

项目学科： 计算机科学学科

项目作者： 郎荣玲

作者单位： 北京航空航天大学

项目金额： 24万元

中文摘要： 数据驱动的飞机故障诊断技术可以充分利用飞行数据对飞机进行故障诊断，提高飞机的安全性。飞行数据的复杂性、多维性、随机性、不完整性及不平衡性等特点，造成故障诊断结果对具体飞行数据值以及随机噪声敏感等问题。本项目主要研究如何提高数据驱动的飞机故障诊断技术的鲁棒性，并对设计的方案和算法进行理论分析和仿真验证。研究内容主要包括不完整飞行数据源的完备化处理、特征性能参数提取、鲁棒的多维非线性飞机性能参数分类。具体包括提出了鲁棒竞争聚类技术，并利用该技术对无标记飞行数据进行样本化处理；采用相关特征向量机方法获取飞机的性能参数与故障模式的关联关系；建立了摄动支持向量机模型，并利用该模型进行故障诊断，提高了诊断算法对噪声的免疫能力。力争通过本项目的研究，提高数据驱动的故障诊断系统在各种噪声扰动情形下保持工作性能指标的能力。为我国军机和民机的故障诊断、故障预报、健康管理等技术的发展提供可行的解决途径。

中文关键词： 故障诊断；数据驱动；鲁棒性；；

英文摘要： The technologies of data-driven fault diagnosis are important for airplane to ensure its security by using flight data. Flight data has some characters, such as complexity, multidimensionality, randomicity, incompleteness and imbalance, which result in the diagnostic results being sensitive to the specific values and random noise. In this project, we mainly study how to improve the robustness of data-driven fault diagnosis for airplane, and we also analyze and validate the proposed approaches. This project addresses three major issues, which are completion of incomplete flight data, selection of the performance parameters and classification of the multidimensional and nonlinear performance parameters with robustness. A robust competitive clustering algorithm is proposed to classify the unmarked flight data, which alleviate the incompleteness of the flight data. The relationships between the performance parameters and fault modes are obtained by relevance feature vector machine. A model of support vector machine with perturbation is proposed in this project. The model can not only be used for classifing the performance parameters, but also improve the diagnostic result's immunity to noise. We try to improve the data-driven fault diagnosis's ability of keeping the performance under all conditions by studying the

英文关键词： fault diagnosis；data driven；robustness；；