Table 1.
AlSi10Mg powder composition.
Table 2.
LPBF processing parameters used and their respective volumetric energy density.
Fig 1.
a) Specimen printed using LPBF process, and b) Tensile test specimen.
Fig 2.
Plots of Gaussian processes considering variables x1 and x2, a) 2-D contour plot of GP, and b) 3-D Surface plot of GP.
Fig 3.
Normal distribution of the RBF Kernel.
Fig 4.
Visualization of tree structure in graphical plot.
Fig 5.
2D visualization of tree structure consisting of data splits with respect to each feature such as (a) VED vs UTS, (b) Hatch Distance vs UTS, (c) Laser power vs UTS, (d) Layer thickness vs UTS, and (e) Exposure time vs UTS.
Fig 6.
3D visualizations of tree structure considering 2 features such as (a) Laser power & VED, and (b) Exposure time & VED.
Table 3.
Various tensile properties of heat treated and non-heat treated AlSi10Mg samples.
Table 4.
Values of UTS prediction performance indicators for heat treated samples.
Table 5.
Values of UTS prediction performance indicators for non heat-treated samples.
Fig 7.
Scatter plots for UTS prediction for heat treated and non-heat treated samples.
Scatter plots are used in MLR algorithms to illustrate the correlation between predicted and actual values. This helps in assessing the efficacy of the model. Scatter plots typically display the x-axis as the actual values and the y-axis as the predicted values. Optimally, the data points ought to align precisely along a 45-degree diagonal line, signifying precise predictions where the anticipated values correspond closely to the actual values. Deviations from the above approach differentiate between model forecasts and actual data, emphasizing regions of excessive fitting or inadequate fitting. Scatter plots can be used to measure model accuracy, bias, variance, and overall goodness of fit, making them a crucial tool for evaluating and improving regression models [46]. Fig 7 illustrates the scatter diagrams produced by different MLR models used in this investigation to visually represent the relationship between the actual and predicted UTS of both heat-treated and non-heat-treated AlSi10Mg samples.
Table 6.
Values of yield strength prediction performance indicators of heat treated samples.
Table 7.
Values of yield strength prediction performance indicators of non-heat treated samples.
Fig 8.
Scatter plots for yield strength prediction for heat treated and non-heat treated samples.
The scatter plots generated by the machine learning regression models employed in this study to represent the correlation between the observed and estimated the yield strength of both heat treated and non-heat treated AlSi10Mg samples are illustrated in Fig 8. This plot illustrates the relationship of the projected values with the empirical data for each of the four models. Scatter plots demonstrate that non-linear models, such as gaussian process regression, random forest regression, and decision trees, outperform LR in terms of the accuracy of non-heat treated samples.
Table 8.
Values of percentage elongation prediction performance indicators of heat treated samples.
Table 9.
Values of percentage elongation prediction performance indicators of non-heat treated samples.
Fig 9.
Scatter plots for percentage elongation prediction for heat treated and non-heat treated samples.