The powder bed fusion (PBF) process is governed by a complex interplay of kinetic and thermodynamic mechanisms that occur across multiple spatial and temporal scales. Consequently, the process exhibits highly dynamic behavior, where even minor variations in boundary conditions can lead to locally altered melting conditions and process anomalies. These anomalies manifest as inhomogeneous material properties and defects within the manufactured parts, reducing process reproducibility and part quality.

Previous research has extensively investigated the challenge of locally varying process conditions. Existing approaches to laser power control and regulation aim to homogenize the energy input using melt pool models or process monitoring data. However, most studies focus primarily on geometry-based influences and validate their results using geometric parameters, without a comprehensive assessment of the resulting melt pool behavior or local material properties. Moreover, experiments are often limited to single-system setups, hindering the transferability of findings across different machines, materials, and monitoring configurations.

This study addresses the research question: How can a data-driven, locally adaptive process strategy correct process influences to reduce process anomalies and improve part properties? Two predictive approaches are developed: one based on a dimensionless, physically derived characteristic parameter, and another utilizing process monitoring data for relative comparison with a calibrated reference process. Both are validated through deviations in melt track geometry and fatigue strength, and subsequently tested on a different PBF system to assess general applicability.

The implementation of adaptive process parameters achieves a system-independent homogenization of melt track geometry while maintaining consistent 50% fatigue strength for alloy 316L. The proposed methodology provides an easily transferable framework for industry applications, enabling local process adaptation without requiring extensive domain knowledge. Furthermore, it allows the formation of controlled microstructures and offers potential for realizing graded material properties through targeted parameter modification.