PSI - Issue 83

C. Mallor et al. / Procedia Structural Integrity 83 (2026) 130–137

133

Ex-situ characterization centred on distortion after partial release from the plate. Each build contained four twin cantilevers, which were first separated from one another using an oscillating saw while still retaining the central attachment to a local portion of build-plate material. Measurements were then taken along the X-Z centre plane at evenly spaced positions on the top surface of the cantilever as shown in Fig. 4. A calliper and an optical microscope were used to record the vertical height before cutting and after partial cutting.

Fig. 4. Vertical deflection measurement points on the top surface of the twin-cantilever along the centre plane for ex-situ distortion measurement. The distortion metric used throughout the paper is the difference between of the vertical displacements after and before the partial cut as shown in Eq. (1), where ௜ refers to the vertical deflection of point , measured at the same location. This quantity is the key link between experiment, simulation, and surrogate prediction. ௜ ൌ ݖ ௜ ௔௙௧௘௥ െ ݖ ௜ ௕௘௙௢௥௘ (1) 2.2. Simulation Strategy The high fidelity model developed is grounded in a physics-based framework implemented in Abaqus [26], which provides a state-of-the-art simulation strategy for additive manufacturing. The underlying physics-based methodology is based on a sequentially coupled thermo-mechanical analysis. Geometry is created in CAD, slicing data is converted into a machine toolpath, and that toolpath is translated into an event-series containing time, spatial coordinates, and process parameters. Material activation and heat input are then applied progressively according to the build sequence. This makes possible to reproduce the order of supports, legs, and cantilever layers, as well as the pauses between scans. The FEM-based approach offers a comprehensive simulation of the temporal and spatial distributions of thermo-mechanical variables, including temperature, deformation, strain, and stress, which are related to the mechanical behavior of the component. Laser heating was represented with a moving Goldak double-ellipsoidal heat source [27] that travelled along the scan path. The double ellipsoidal Goldak heat source model is presented in Fig. 5.

Fig. 5. Goldak model for double ellipsoid heat  ux distribution in the PBF-LB process.

Temperature-dependent properties for 316L stainless steel included density, conductivity, specific heat, coefficient of thermal expansion, elastic modulus and Poisson ratio, obtained from literature [28–30] and collected in [23]. The advantage of this framework is that it is parameterized with the same process variables later used in the design of experiments: laser power, scan speed, and build-plate preheating. Since those inputs have a direct physical meaning, the resulting high fidelity model can be sampled to create a dataspace to build a surrogate. The simulation framework therefore serves both as a validation target and as a data generator for the reduced order model (ROM).