A morphology-aware process–structure–property framework for SLM AlSi10Mg: Decoupling grain and dendrite contributions

• Develop an automated multi-scale PSP framework for AM process optimization. • Integrate ML and physics simulation to build an interpretable process map. • Separate microstructure predictors from misleading proxies for performance. • Identify equiaxed fraction as the dominant yield strength predictor. • Confirm that dendrite morphology is a misleading proxy for hardness. Establishing a robust process–structure–property (PSP) framework is critical for designing additively manufactured (AM) alloys with tailored performance. Here we introduce a morphology-aware PSP workflow for SLM AlSi10Mg that, for the first time, enables systematic quantification of dendritic/cellular morphologies by reconstructing the Si network into closed cells. This high-throughput approach, combined with EBSD grain analysis, thermophysical modelling, and validated statistics, allows direct comparison of grain- and dendrite-level predictors. We show that grain classification (equiaxed vs. columnar) refines the Hall–Petch relation and reliably predicts yield strength. In contrast, while dendritic morphology is widely assumed to correlate with hardness, it fails as a predictor. Moreover, this work presents direct microstructural evidence for decoupling between grains and dendrites. For instance, in the melt pool boundary region, columnar grains are predominantly filled with equiaxed dendrites. This indicates that the final dendritic morphology is more strongly influenced by local solidification conditions than by the parent grain it resides in. The proposed framework is potentially transferable to other AM alloys with cellular or dendritic microstructures (e.g., SLM Ti-6Al-4V, IN718) and highlights the risk of relying on visually intuitive but statistically unsupported proxies in PSP framework.