CINDERELLA : 3D metal printing control using innovative laser sources, digital augmented models and artificial intelligence.

The CINDERELLA project aims to contribute to accelerating the mastery of the LPBF process by adopting an approach that is both experimental and numerically assisted by Artificial Intelligence (AI), mastering laser-material interaction at the level of the liquid metal bath and microstructure generation. The results of the project will contribute to: (1) accelerating parametric optimization for new materials and materials known to be difficult, (2) to obtaining the desired microstructures at the end of the process, coupled with the minimization of thermal stresses/deformations in components, thereby reducing the need for part support and, potentially, enabling process control, (3) to the development of more efficient numerical models, integrating the variety of different study scales.

To this end, the project proposes a two-pronged approach. The first is advanced experimentation with the process: Previous studies have demonstrated the benefits of spatial and temporal shaping of laser beams for processing difficult materials, adapting local microstructures to specific requirements, and reducing porosity. Instrumented experiments will explore the effects of variations in the configurations of this type of laser source.

AI tools will be used to optimize parametric research (assisted experimental designs) and establish phenomenological models that can be used for coupling with finite element (FE) numerical models and to refine online process control. Advanced instrumentation and data processing will accelerate the identification of these models: high-speed cameras for studying the melt/solidification bath, rapid determination of microstructural information using ultrasonic sensors. The second focus is on AI-assisted multiphysics modeling/simulation of the LPBF process. The project aims to ultimately develop a digital twin of the process. The approach proposed by CINDERELLA consists of coupling phenomenological models obtained by processing experimental data using AI with FE models (model reduction using AI) and cellular automata (CAFE) in order to significantly reduce calculation times and improve model response.

These numerical models take into account laser-material interaction, the melt pool, and microstructure generation, drawing on the development of microstructural prediction approaches based on cellular automata methods optimized by coupling with phenomenological models. The new numerical strategies will first be validated on well-understood materials by comparing the simulation results with experimental data over a wide range of parameters (study of model robustness and generalization). Subsequently, the capabilities of the project’s experimental and numerical developments will be evaluated on materials that are currently more difficult to use in LPBF.