Introduction: Common defects in printed metallic components affect the mechanical properties and serviceability of parts. For example, lack of fusion voids and solidification cracking are known to adversely affect tensile properties. Warping, buckling and delamination of components because of high residual stresses cause dimensional inaccuracy, and in extreme cases, part rejection. The formation of these defects depends on multiple complex physical processes and there is no simple method to mitigate them based on scientific principles. Here, we identify conditions for minimizing lack of fusion voids, solidification cracking, and delamination in printed metallic parts by using a synergistic combination of a mechanistic model, machine learning and experimental data on defects.
Technical approach: We collect independent experimental data on defect formation from the literature for stainless steel, nickel, titanium and aluminum alloys. The data are used to train, validate and test a neural network, a random forest, and a decision tree. First, the alloy properties and printing process parameters such as laser power, scanning speed, layer thickness, hatch spacing, preheat temperature, and substrate thickness are used to forecast the occurrence of lack of fusion voids, solidification cracking and delamination using neural networks. Then the hierarchical influences of several complex variables on the defect formations are predicted using random forests and decision trees. These complex variables include peak temperature, molten pool dimensions, substrate rigidity, cooling rate during solidification, solidification parameters, Fourier number, and Marangoni number. The variables are calculated using a well-tested heat transfer and fluid flow model. The model solves the equations of conservations of mass, momentum, and energy in a 3-D discretized solution domain that includes deposit, substrate, and shielding gas. Thermophysical properties of stainless steel, nickel, titanium and aluminum alloys required for the heat transfer and fluid flow model are computed by thermodynamic calculations using a commercial software JMatPro.
Results and discussions: Temperature gradient during cooling, molten pool dimensions and cooling rate during solidification are the most important variables that control delamination, lack of fusion voids and solidification cracking. Both the random forest and decision tree could forecast the defect formation better than the neural network. Predictions of defects using the computed variables as inputs are more accurate than that using the raw printing process parameters.
Summary and conclusions: In summary, a synergistic combination of heat transfer and fluid flow model of metal printing, machine learning algorithms such as neural network, decision tree, and random forest and experimental data is used to predict conditions to mitigate defects in printed parts. The computed variables are superior to the raw printing process parameters in forecasting defect formation.