Process development for metal additive manufacturing by laser powder bed fusion (LPBF) aims to find a stable parameter set with minimal effort of time and resources but maximal performance and reliability. The multistage process includes first achieving general processability and high density, which acts a major target for most materials. Afterwards, further properties such as built rate, mechanical behaviour, geometrical accuracy and surface quality are treated, which may require further modifications in the manufacturing strategy.

The complex multiscale and multiparametric nature of LPBF processing makes it challenging to predict the materials behaviour during manufacturing precisely. Therefore, only a prognosis of a suited parameter range is possible based on valid preknowledge.

Our contribution covers the question, how a suited design of experiment (DOE) for LPBF process development should be constructed and what insights on the impact of several major LPBF-parameter and their possible interactions on a target value can be gained. For this reason, two different DOE configurations are applied to the process development of the material Invar/FeNi36: A D-optimal layout with 18 samples and a full factorial layout with 64 samples, whereby both describe the same parameter space.

The impact of laser power, scan velocity and hatch spacing can be clearly separated in both cases, if the experimental variety is large and a sufficient prediction of part density can be achieved. If the parameter space becomes too narrow, a clear prediction of an optimal parameter set becomes highly inaccurate, whereby the general trends in dependencies are still preserved. Furthermore, the reliability of the results is treated and the process-related scattering is discussed.