The past two decades witnessed the evolution of non-stationary thermal wave imaging (NSTWI) techniques for active infrared non-destructive testing of various industrial components. The non-stationary temporal thermal response from the test sample in NSTWI comprises of different components, where a proper signal decomposition algorithm decomposes all these components and provides actual thermal response of the test object for defect detection. The present article qualitatively analyses the three signal decomposition algorithms such as empirical mode decomposition (EMD), Hilbert vibrational decomposition (HVD) and variational mode decomposition (VMD) for thermal signal decomposition and defect detection. A mild steel specimen with artificially simulated defects of same size lying at various depths is used to experimentally validate the three signal decomposition algorithms. Further, the defect detection is carried out by employing Fourier transform phase on each intrinsic mode function (IMF) of the three decomposition algorithms. Defect signal-to-noise ratio is considered for qualitative comparison of three signal decomposition techniques in NSTWI.

过去的二十年见证了用于对各种工业部件进行主动红外无损检测的非静止热波成像 (NSTWI) 技术的发展。NSTWI 中测试样品的非平稳时间热响应由不同的组件组成，其中适当的信号分解算法分解所有这些组件并提供测试对象的实际热响应以进行缺陷检测。本文定性分析了经验模态分解（EMD）、希尔伯特振动分解（HVD）和变分模态分解（VMD）等三种信号分解算法用于热信号分解和缺陷检测。使用具有不同深度的相同尺寸的人工模拟缺陷的低碳钢样品对三种信号分解算法进行了实验验证。此外，通过对三种分解算法的每个固有模式函数（IMF）采用傅里叶变换相位来进行缺陷检测。NSTWI中三种信号分解技术的定性比较考虑了缺陷信噪比。