Enhanced pre-processing of thermal data in long pulse thermography using the Levenberg-Marquardt algorithm

Long pulse thermography (LPT) has recently been proposed as a new promising and cost-effective active thermographic technique over traditional flash thermography. However, LPT can be affected by uneven heating noise when two or multiple optical heat excitation sources are used. Signal pre-progressing algorithms such as the background subtraction method are typically used to remove uneven heating noise. In this study, a novel signal pre-progressing approach to remove uneven heating noise for LPT is presented, which relies on the Levenberg-Marquardt (LM) algorithm. Unlike the Ordinary Least Squares (OLS) method, LM operates an iterative optimization process to fitting the raw thermal data that allows selecting thermal signals contained only in the sound area. Results showed that the LM algorithm provided higher efficiency than the OLS algorithm to denoise thermal data. Moreover, the LM method combined with principal component analysis further enhanced the capability of LPT to visualize material damage.