Numerical-assisted prediction model of layer height for Co-Cr-Ni-alloy direct energy deposition

During the L-DED multilayer stacking process, the molten pool dynamic behavior and liquid metal flow phenomenon still need to be clarified, hindering the optimization of cladding formation quality. Here, the molten pool overflow (MPO) phenomenon is defined from a theoretical analysis model and described through multi-source information sensing. We conduct a three-dimensional numerical model to reveal that MPO is due to the temperature and geometry fluctuations of the molten pool. MPO is also experimentally visualized using temperature and image data, and its geometric properties are determined. Hull depth, length, and distance can quantitatively describe the MPO phenomenon. Then we propose a numerical-assisted RF-LSTM (Random Forest and Long Short-Term Memory) prediction model for cladding layer height; combining the numerical model with machine learning can improve the model’s prediction accuracy. The proposed model offers insights into understanding and predicting cladding layer height variations in additive manufacturing processes.