论文部分内容阅读
针对当前结构健康监测系统中信号数据量过大、所占内存过高所导致的难以传输和储存的问题,提出将压缩感知方法运用于神经网络系统与压电阻抗(Electro-mechanical impedance,简称EMI)技术进行结构的损伤识别。以压缩后的EMI数据代替原始数据来表征原始损伤信号,用离散余弦变换分析了EMI信号的稀疏度,并以高斯随机矩阵作为观测矩阵,满足了压缩感知理论的限制性等距条件。以齿轮结构为例,采用有限元技术构建了EMI模型,得到了不同裂纹深度下的电导(电阻抗的倒数)信号。采用压缩感知方法先对电导信号进行压缩,再进行主成分分析,以降低BP神经网络输入参数的维度。结果表明:使用压缩感知技术之后,数据传输带宽和储存空间只需为原来的44%;将压缩后的电导数据的主成分作为神经网络的输入参数,该系统能够检测到结构裂纹的存在,并能对裂纹的损伤程度进行有效的定量归类。
In order to solve the problem that it is hard to transmit and store due to too much data volume and high memory in the current structural health monitoring system, a compressive sensing method is proposed to apply to the neural network system and the electro-mechanical impedance ) Technology for structural damage identification. The original damage signal is represented by the compressed EMI data instead of the original data. Discrete cosine transform is used to analyze the sparsity of the EMI signal. The Gaussian random matrix is used as the observation matrix to satisfy the constrained isometric condition of compressed sensing theory. Taking the gear structure as an example, the EMI model was constructed by using the finite element method and the conductance (reciprocal of the electrical impedance) signals at different crack depths were obtained. Compressive sensing method is used to compress the conductance signal first, and then the principal component analysis is performed to reduce the dimension of input parameters of BP neural network. The results show that using compressed sensing technology, the data transmission bandwidth and storage space need only be 44% of the original. The main component of the compressed conductance data is used as the input parameter of the neural network. The system can detect the existence of structural cracks and It can effectively classify the degree of damage to the crack.