Bayesian optimization for extreme high-speed laser material deposition

Extreme high-speed laser material deposition, also known by the acronym EHLA, enables metallic coatings of different thicknesses at deposition speeds of up to several hundred meters per minute and deposition rates of several kilograms per hour. Against other deposition welding processes, EHLA offers significant advantages in terms of lower heat input and higher precision, making it a valuable option for processing materials, which are considered hard-to-weld. Despite its advantages, the highly nonlinear interdependencies of multiple influencing variables require precise control and tuning of the parameters and challenge the process development, making it time and cost expensive. In the absence of an accurate process model for large parameter spaces, model-based optimization is currently not feasible, such that current development requires extensive experimentation and expert knowledge. To overcome these challenges, an adaptive process development approach for the key process parameters, such as laser power and powder flow rate, based on Bayesian optimization (BO) is proposed. BO employs probabilistic models trained on experimental data to systematically explore the parameter space and predict the optimal settings in terms of a target variable. The investigations show that the sample-efficient, data-driven method effectively accelerates the development of suitable process parameters and drastically reduces the need for extensive empirical testing and expert knowledge.