Abstract

Characterization of various industrial components without impairing their future utility increases the necessity of nondestructive testing (NDT) techniques. In recent years, active thermography catches the researcher’s interest due to its distinct abilities like whole-field, contact-free, economic, defended, noninvasive, and nondestructive type of investigation. High energy deposition at low frequencies accompanied with enhanced depth resolution requirements of active thermography grabs the interest of quadratic frequency-modulated (QFM) stimulus in the recent past. On the other hand, enhanced defect detection is dependent on the extraction of an appropriate time-frequency component of chirped thermal response through a suitable processing technique. The present work employs velocity synchronous linear chirplet transformation (VSLCT) to precisely map the nonlinear chirp rate of the extracted quadratic frequency modulated thermal response. The proposed methodology is validated by carrying experimentation over carbon fiber and glass fiber reinforced polymer samples with artificially made flat bottom holes and Teflon inserts. Defect detectability of the proposed technique is quantified by comparing conventional techniques with the figure of merit (FoM) as signal to noise ratio (SNR), full width at half maxima (FWHM), and probability of detection (POD). An arduous manual inspection of processed data recommends automatic defect detection through segmentation without manual intervention. An active contour-based segmentation is performed post to VSLCT technique to facilitate the automatic visualization of defects.

中文翻译：

---

使用非平稳热波成像的复合材料缺陷检测方法

摘要

在不影响其未来效用的情况下对各种工业部件进行表征增加了无损检测 (NDT) 技术的必要性。近年来，主动热成像技术以其全场、非接触、经济、防御、非侵入性和非破坏性调查等独特能力引起了研究人员的兴趣。低频下的高能量沉积以及主动热成像对深度分辨率的要求提高，最近引起了二次调频 (QFM) 刺激的兴趣。另一方面，增强型缺陷检测取决于通过合适的处理技术提取啁啾热响应的合适时频分量。目前的工作采用速度同步线性啁啾变换 (VSLCT) 来精确映射提取的二次调频热响应的非线性啁啾率。所提出的方法通过对碳纤维和玻璃纤维增​​强聚合物样品进行实验来验证，这些样品具有人工制作的平底孔和特氟龙插件。通过将传统技术与作为信噪比 (SNR)、半峰全宽 (FWHM) 和检测概率 (POD) 的品质因数 (FoM) 进行比较，可以量化所提出技术的缺陷可检测性。对处理过的数据进行繁重的人工检查，建议通过分割进行自动缺陷检测，无需人工干预。