高铁钢轨表面缺陷的光声无损检测方法研究

The research proposed photoacoustic nondestructive detecting methods on high-speed rail surface defects, including three aspects: the modeling of surface acoustic wave producing by the laser detection of rail surface defects, the establishment of photoacoustic imaging of real-time non-destructive testing systems based on array detector, photoacoustic signal feature extration and classification of rail surface defects. By photoacoustic non-destructive testing technique, photoacoustic signal of the rail surface excitated by the pulse laser was measured and preprocessed, and then the signal and the reconstructed image were decomposed by using Hilbert-Huang transform to analyze basic amplitude-frequency characteristics. Multi-dimensional tensor was composed by the photoacoustic signal frequency, amplitude and different measurement points, then second feature was extracted by using non-negative tensor decomposition to get the injury characteristic coefficient and characterization information of material variances. At the same time, photoacoustic signal characteristic coefficients in the case of defectfree and typical defects were obtained by the bulit wave model, the defect recognition rules were established by using the the relevance vector machine classification features. Finally, real-time defect detection of the measured photoacoustic signal was achieved by the establishment of the defect identification rules. All in all, the research proposed a new track surface defect detection method, which can accurately identify the implicit rail defect information in real time, effectively ensure high-speed rail operation safely.

该项目提出了高铁钢轨表面缺陷的光声无损检测方法，包括激光检测钢轨表面缺陷产生声表面波的建模、基于阵列探测器的光声成像实时无损检测系统的建立和钢轨表面缺陷光声信号的特征提取及分类三方面的内容。通过光声无损检测技术测量激光激发钢轨表面的光声信号，并对信号进行预处理，利用希尔伯特-黄变换对信号和重建图像进行基本幅频特征分析；然后，结合光声信号的频率、幅值以及不同的测量点组成多维张量，利用非负张量分解进行二次特征提取，得到缺陷特征系数和表征材料差异的特征信息；同时，结合激光检测钢轨表面缺陷产生声表面波的模型，获得在无缺陷以及典型缺陷情况下光声信号的特征系数，利用相关向量机对特征进行分类并建立缺陷识别规则；最后，通过建立的缺陷识别规则，实现对实测信号的缺陷实时检测。该项研究作为一种全新的轨道表面缺陷检测方法，能准确、实时地识别钢轨中的隐含伤损信息，保证高铁安全有效的运行。

本项目以利用光声信号检测到高铁钢轨表面缺陷信息为目标，包括如何获取和处理钢轨表面光声信号、如何对缺陷进行特征提取、分类和识别等，研究要点包括：钢轨表面典型缺陷的光声信号建模与数值求解、光声无损检测系统的搭建、钢轨典型缺陷光声信号的特征提取与分类算法的设计。针对钢轨表面典型缺陷的光声信号建模与数值求解，提出了利用有限元仿真得到初始声压，再利用K-Wave仿真声场传播的方法，得到了几种典型缺陷光声信号的数值解。建立了钢轨内部夹渣缺陷的模型，并通过仿真研究了利用光声弹性特征检测钢轨内部缺陷的方法。搭建了钢轨缺陷光声无损检测系统，并针对光声信号中噪声的非平稳和非线性特点，设计了一种基于EMD-Energy的光声信号去噪算法，系统的检测精度达到0.1mm，检测深度达到5mm。针对钢轨典型缺陷光声信号，分别设计了基于希尔伯特黄变换和支持向量机(HHT+SVM)的方法以及基于非负张量分解和极限学习机(NTF+ELM)的方法进行特征提取、分类和识别，分类准确率都超过了98%。此外，本项目在还进行了钢轨缺陷多模态信号无损检测方法的研究，设计了多模态无损检测方法和系统，可实现对钢轨缺陷的快速、全面的检测。