Mathematical modelling to predict the tensile strength of additively manufactured AlSi10Mg alloy using artificial neural networks

Integrating machine learning in additive manufacturing to simulate real manufacturing outcomes can significantly reduce the cost of manufacturing through selective manufacturing. However, limited research exists on developing a prediction model for the mechanical properties of the material. The input variables include key selective laser melting process parameters such as laser power, layer thickness, scan speed, and hatch spacing, with tensile strength as the output. The artificial neural network (ANN) based mathematical model is compared with a second-degree polynomial regression model. The robustness of both models was further assessed with the new data points beyond those used in the development of ANN-based mathematical model and regression model. The results demonstrate that the proposed ANN-based mathematical model offers superior accuracy, with a mean absolute percentage error (MAPE) value of 4.74 % and the R2 (goodness of fit) value of 0.898 in predicting the strength of AlSi10Mg. The ANN-based mathematical method also demonstrates the strong performance on the new data, achieving a regression value of 0.68. This concludes that the model shows sufficient proof to consider a viable option for predicting the tensile strength.