Recently, online process monitoring has become an essential part in industrialized laser-based surface pre-treatment applications to enhance both efficiency and process stability while ensuring high-quality processed parts and reducing the production of scrap parts. In this study, aluminum parts are processed with a nanosecond pulsed laser to remove contaminations and to produce an appropriate surface topography that ensures good adhesion between the aluminum parts in a subsequent bonding process. A diode-based monitoring system with a high temporal resolution is implemented and used to detect electromagnetic radiations generated by the laser process on aluminum surfaces with different initial roughness and contamination state. Moreover, machine learning models are developed for evaluation of the obtained data. The output of these models is the success probability of the surface functionalization which is determined by the adhesive properties of the corresponding surfaces.

The results show that the developed models were able to cluster the resulting laser-treated surfaces according to their initial state (e.g. roughness or presence of paint or oil). Then a linked model uses this initial prediction of the surface state to enhance the prediction of the adhesion properties. It can be concluded that the machine learning models successfully enabled the development of a high temporal resolution monitoring method for an industrialized laser-based pre-treatment process e.g. in the main processing line of the battery production for electro mobility.