Predicting laser-induced surface modifications on metals in laser surface processing using a hybrid deep learning approach

Laser-induced surface modification (LISM) is essential for precise surface functionalization in laser surface processing (LSP). This study examines the performance of a hybrid deep learning model, which combines a convolutional neural network (CNN) with a multi-layer perceptron (MLP), to enhance the accuracy of predicting laser-induced surface modification (LISM) on various metals. The CNN is designed to extract the intricate features from laser-irradiated surface images (LISI) of metals, while the MLP processes numerical and categorical values of laser intensity to learn the complex relationships. The proposed CNN-MLP model was trained and validated on a diverse dataset generated by nanosecond Nd: YAG pulsed laser irradiation at varying energy levels. The model demonstrated superior performance, achieving an accuracy of over 99% in predicting LISM on metals, whether modified or unmodified, and outperforming standalone deep learning models in terms of generalization and predictive reliability. This research presents an effective pathway toward intelligent LSP systems, supporting process automation, enhanced quality assurance, and significantly less reliance on trial-and-error experimentation.