Explainable defect detection and Recipe recommendation for wood-based panels

This nugget is developed by the University of Porto. It is about explainable defect detection and recipe recommendation for laser seam welding. This nugget has the same production example as nugget 6; again laser seam welding. The learning goals are to gain additional insights from the predictions produced by a model, understand its recommendations and leverage this knowledge by finetuning and perfecting the real time productions. Explainable AI (XAI) is introduced. For laser seam welding, the main process conditions are the metal sheet thicknesses. The goal is to find a set of parameters that influence the quality of the welding. The parameters used are input for a parameter explainable model. The analysis of the parameters of the laser welding show that width and depth are directly proportional and highly correlated. The parameter optimization process is done by the Basin-Hopping algorithm.