Time-efficient surrogate models of thermal modeling in laser powder bed fusion

Two time-efficient surrogate models are proposed to emulate the nonlinear heat equation in context of laser powder bed fusion, performance which is compared accuracy and online execution time. Fast-computed numerical solvers critical implementing digital twin framework additive manufacturing process addressing one its main open problems: lack quality assurance. The first model reduced Gaussian emulator. It a data-driven equipped with dimension reduction scheme manages predict temperature profiles almost instantly (around 0.036s on average) an 95% for 99.38% tests. Another sketched emulator local projection. projects accurate but high-dimensional finite element method solution low-dimensional basis then bypasses majority costly computations temperature-dependent matrices projected by randomized sketching. has higher (97.78% tests relative errors below 1%) while spending comparably more time 42.23s average). Although both surrogates promote efficiency some minor controlled compromise accuracy, enables real-time implementation projection offers levels accuracy. A series experiments carried out, assumes three-layer printing fixed beam trajectory using small number control parameters as inputs, namely power, scan speed, coordinates. Both also principally feasible other thermal-driven obtain better assurance techniques like uncertainty management closed-loop control.