Towards accurate thermal simulations in PBF : a novel calibration strategy for Gauss-Goldak heat-source models

Laser Powder Bed Fusion (PBF-LB) is an advanced additive manufacturing process that fabricates near-net-shape metal components by selectively melting and solidifying powder layers with a laser beam. The intense, localized energy input generates complex thermal cycles that strongly influence microstructural evolution, residual stresses, and mechanical performance. Accurate prediction and control of this thermal history are therefore critical for optimizing part quality and process reliability. Finite Element (FE) models employing the Gauss-Goldak heat source are widely used to simulate melt pool temperature fields and support residual stress analyses. However, the accuracy of these requires precise calibration of the heat source parameters, particularly the laser absorptivity and volumetric heat flux distribution, which vary with process conditions. This study introduces a novel calibration approach combining melt pool geometry data and in-situ thermographic measurements from the bottom side of a substrate with thermal simulations to inversely identify these parameters. The method enables tailored tuning of the thermal heat input model for specific processing conditions, substantially improving the accuracy of transient temperature predictions. Bayesian optimization was used to effectively identify absorptivity from the far-field temperature, and the shape parameters of the Gauss-Goldak heat source from the melt pool geometry. The calibrated model demonstrates excellent agreement with experimental data, achieving deviations in simulated melt pool dimensions below 2.8%. Thereby, absorptivity emerges as the key parameter governing the far-field thermal response, underscoring the importance of its precise calibration for accurate part-scale thermal simulations. Furthermore, it is shown that, within the investigated parameter space, the shape parameters of the Goldak model exert negligible influence on the far-field temperature distribution, justifying a two-step approach. Although the calibration was performed using the Gauss-Goldak model, the approach is generalizable and can be extended to alternative heat input formulations. This enhanced modeling fidelity enabled by an accurately calibrated heat-source can facilitate more reliable residual stress and distortion forecasts in PBF manufacturing.