AI Model Finds Viable Low-Power Parameters for GRCop-42 3D Printing After Screening 100 Million Configurations

The team trained a probabilistic model on prior failed builds to predict print success across a 100-million-point parameter space of laser power, scan speed, and hatch spacing. It then iteratively selected batches that balanced exploitation of promising regions with exploration of high-uncertainty zones, feeding results from physical prints back into the model each round. This reduced the experimental burden from an estimated 100 million trials to a few dozen targeted prints. The approach directly addresses the high cost and material waste that had confined GRCop-42 to specialized high-power systems. By succeeding at lower energies, the method opens the alloy for broader aerospace and heat-exchanger applications on standard commercial hardware. The same Bayesian optimization framework can transfer to other high-dimensional materials problems such as alloy design or drug formulation where exhaustive testing is prohibitive.