Exploiting data science for atmospheric plasma enhanced paint adhesion on carbon

Abstract

Carbon fiber-based thermosets, or thermoplastic composites, used in automotive and aerospace industry are prone to environmental degradation and often require a protective coating. These composites typically have low surface energy and are difficult for any protective coating to adhere to their surface. Plasma, the fourth state of matter, can be used as a surface modification method that initiates formation of oxygen and nitrogen based free radicals, ions, and electrons, activates the surface molecules, and promotes paint or coating adhesion. In this study, a commercially available atmospheric pressure plasma jet (APPJ) operating with clean, dry air (CDA) was used for the surface treatment of AS4/3501-6 composite laminate coupons. APPJs are used as highly scalable, energy efficient, and generate no hazardous waste. However, optimization of plasma-based processes generally requires multi-parameter and multi-disciplinary testing for best results, which could be time consuming. Minimizing the experimental data collection for optimization of functional properties can be accomplished through data science. Bayesian optimization (BO) was used to enhance paint adhesion by training a model based on the percentage area intact after paint adhesion test (ASTM 3359-23 Method B). Empirical data for the optimization was fed into the model from a short design of experiments based on 3-level full factorial method varying treatment parameters: number of passes and speed of plasma jet, and stand-off distance. Results showed that using BO, the untreated material adhesion area improved to over 98% paint adhesion with optimized APPJ treatment parameters. The surfaces were further characterized and correlated to paint adhesion via changes in surface roughness through optical profilometry, in water contact angle goniometry, and in chemical composition through X-ray Photoelectron Spectroscopy (XPS analysis). For example, the water contact angle decreased from the untreated, 72°, to empirical best at 28°, and further to 15° from the optimized parameters explored using BO. Bayesian optimization’s ability to adapt and fine-tune the plasma treatment parameters makes it a powerful tool in the quest for discovering and optimizing the adhesion performance of paints on carbon fiber reinforced composites, contributing to improved performance.