Computation of tensile and compressive strengths of additively manufactured ABS material for automotive applications using ANN algorithms

Experimental approaches and finite element analysis (FEA) are used to optimise the material extrusion (MEX) parameters for achieving effective tensile and compressive strengths of acrylonitrile-butadiene-styrene (ABS), which can lead to material waste and be costly. Additionally, techniques such as basic empirical models have failed to capture the complex non-linear interactions, resulting in inaccurate estimations. To address these challenges and compute the tensile and compressive strengths of additively manufactured ABS, this study develops novel artificial neural network (ANN) models using the Adam-optimised and Bayesian-regularised algorithms trained on an experimental dataset. Compared to traditional benchmarks, this study exhibited exceptionally low mean square error (MSE), root mean square error (RMSE), mean absolute error (MAE) and mean absolute percentage error (MAPE) across hold-out and five-fold cross-validations. Industrial validations using the Taguchi L15 parametric combinations, i.e., print speed, layer height, infill density, bed temperature and nozzle temperature, yielded a coefficient of determination (R²) above 0.90 for both properties. The case studies on brake-pedal and door-handles exhibited that the Adam-optimised and Bayesian-regularised ANN models predicted properties with MAPE below 1 %, reducing material wastage, development and lead times by 70 %, 60 % and 80 %, respectively. This study indicated the infill density as a critical factor influencing the strengths of the ABS material and thus, contributing towards achieving sustainable development goal (SDG) 9 by ensuring these cost-effective frameworks achieve effective strengths through optimised MEX-print parameters in additive manufacturing (AM) and automotive applications.