Multi-physics modeling and Gaussian process regression analysis of cladding track geometry for direct energy deposition.

• A powder-scale multi-physics model using the Finite Volume Method is developed to study the direct energy deposition process. • The model can simulate cladding track geometry rapidly and accurately. • The influences of the process parameters on the track geometry are analysed in detail using an analysis of variance method. • A Gaussian process regression model is developed to predict the cladding track geometry based on the simulation results. Direct energy deposition (DED) is an effective method to fabricate complex metal thin-wall structures. The geometrical dimensions of the cladding track have significant influence on the dimensional precision of final components. In this study, a powder-scale multi-physics model using the Finite Volume Method (FVM) is developed to study the direct energy deposition process. The mass transfer, phase transformations and heat transfer in the DED process are incorporated and the geometrical characteristics of a single cladding track can be rapidly predicted. The influences of the process parameters including laser power, powder feed rate and scanning speed on the track width and height are analyzed in detail using an analysis of variance (ANOVA) method. Based on the simulation results, a Gaussian process regression (GPR) model is developed to predict the geometrical characteristics of cladding tracks under different manufacturing parameters. Finally, both the multi-physics model and the GPR model are validated by single track deposition experiments. The results show that the proposed multi-physics simulation results are in good agreement with the experimental results and can reveal the qualitative relationship between parameters and track geometry. The GPR model is able to predict the geometrical characteristics of single cladding tracks. [ABSTRACT FROM AUTHOR]