Research on online monitoring of aircraft skin laser paint removal thickness using standard curve method and PCA-SVR based on LIBS

Abstract

High-frequency pulse lasers, applied in the form of rapid scanning, act upon the surface of aircraft skin paint layers, thereby removing the paint layers, exhibiting characteristics of efficiency and eco-friendliness. Real-time monitoring of the paint removal effect and prevention of substrate damage necessitates the continuous monitoring of paint removal thickness. Combining Laser-Induced Breakdown Spectroscopy (LIBS) online monitoring technology enables laser-controlled paint removal under multiple effects coupling, meeting the requirements of airworthiness maintenance. This paper, based on a high-frequency nanosecond infrared pulse laser paint removal LIBS monitoring platform, conducts research on laser paint removal thickness LIBS online monitoring of aluminum alloy plates coated with dual-layer paint. Spectra corresponding to the removal thickness of each group are collected and, respectively, paint removal thickness monitoring models based on LIBS spectra are established using the standard curve method and Principal Component Analysis-Support Vector Regression (PCA-SVR) algorithm. When monitoring paint removal thickness using the standard curve method, the intensity of five Ti element characteristic spectral lines selected is correlated with the paint removal thickness, and segmented curve fitting according to the paint layer structure satisfies the segmented curve fitting of topcoat and topcoat + primer. Among them, the average coefficient of the curve fitting of the Ti II 589.088 nm characteristic spectral line is 0.89, and the root mean square error (RMSE) is 12.28 μm. Its performance is superior in the five standard curves; thus, its fitting equation is used as the criterion for paint removal thickness monitoring. To further improve monitoring accuracy, research on paint removal thickness monitoring models based on PCA-SVR is conducted. Compared to the traditional univariate standard curve method, the PCA-SVR model does not require segmented monitoring. After parameter optimization, the average fitting coefficient reaches 0.97, and the RMSE is 2.92 μm. The results indicate that the PCA-SVR-based paint removal thickness monitoring model has higher accuracy, thereby forming the basis for paint removal thickness monitoring. Through comparative research on paint removal thickness monitoring models, two types of paint removal thickness monitoring criteria are obtained, providing model solutions for high-precision monitoring and automation of aircraft skin laser paint removal thickness.