Color marking of stainless steel has broad applications in decorating souvenirs, jewelries, and creating unique patterns for product tracking and identification. A well-reported approach to achieving a wide range of colors is by controlling the thickness of the laser-induced metal oxide layer. It is, however, challenging to generate a pure color due to the existence of multiple factors that influence the oxide layer thickness. To address this issue, a predictive and reliable model for color generation is highly desirable. In this article, Artificial Neural Network (ANN), as a popular machine learning method, is applied to correlate the relationship between the laser processing parameters and the resulting colors. The experimentally produced RGB (red, green, and blue) colors under various processing parameters of the laser color marking (LCM) are measured and used to train and test the ANN models. The performance of the proposed quantitative model is evaluated by using root mean square errors (RMSE) and correlation coefficients (R). The results show that ANN has a superior ability for predicting the behavior of the LCM process in the generation of RGB colors in terms of the processing parameters (scanning speed and line spacing). Consequently, it can be employed for improving color consistency and repeatability of the process without the trial and error method. As a demonstration, the trained ANN model is practically applied to create RGB images on the surface of stainless steel.
 

laser color marking,artificial neural network,RGB,colorization,Modeling