摘要
结构健康监测是保证航空器持续安全运行的重要方式,正成为无人机平台研发和适航认证的一项关键技术.针对无人机结构动态监测中的多种不同传感器测量信息,实时提取结构加速度、应变响应信号和模态特征参数,构造归一化的小波包能量变化率指标、应变能变化率指标、模态频率变化率指标与混合损伤评价指标,用于指示结构健康状态.利用多层次数据融合技术进行数据级融合、特征融合以及基于Bayes概率神经网络的决策融合,建立结构损伤程度、位置信息与健康评价指标之间的对应关系,通过粗糙集约简显著降低了特征属性的空间维度,获得关于结构健康状况的一致性决策.通过某型号无人机的健康监测实例验证了上述数据融合技术在识别多种类型传感器输入、多位置损伤识别中的精度,表明多元数据融合在无人飞行器结构损伤识别中的有效性.
Structural health monitoring is an important means to guarantee the continuing safe operation of aircrafts,and makes a key technique for unmanned aerial vehicles＇（ UAVs＇） development and certification.For a UAV fuselage,the structural acceleration responses,strain signals and modal parameters were acquired on-line from different sensor measurements in dynamic structure simulation. The normalized wavelet packet energy change rate index,the strain energy change rate index,the modal frequency change rate index and the mixed damage evaluation indices were built to indicate the structural health condition.The integrated multi-source data fusion technique,including data-level fusion,feature-level fusion and Bayesian probabilistic neural network-based decision-level fusion,was used with the rough set reduction successively to significantly decrease the spatial dimension of data. The mapping between structural damage information,like damage severity,damage locations and health evaluation indices,was established,and the comprehensive decision of the structural damage model was achieved. An example for the health monitoring of an unmanned helicopter was demonstrated. The experimental results verify the accuracy of the proposed data fusion technique for damage identification of multi-damage aircraft structures,and show the validity of multi-data fusion in UAV health monitoring.
作者
何绪飞
艾剑良
宋智桃
HE Xufei;AI Jianliang;SONG Zhitao(Department of Aeronautics and Astronautics, Fudan University, Shanghai 200443, P.R. China;Shanghai Aircraft Airworthiness Certification Center of CAAC, Shanghai 200335, P.R. China)
出处
《应用数学和力学》
CSCD
北大核心
2018年第4期395-402,共8页
Applied Mathematics and Mechanics
基金
工信部民机科研专项资助项目(G011605)
中国博士后科学基金(2015M580956)~~
关键词
健康监测
数据融合
概率神经网络
粗糙集约简
损伤识别
health monitoring
data fusion
probabilistic neural network
rough set reduction
damage identification
