Processing and property relationships in material extrusion additive manufacturing of PLA using a hybrid predictive model for tensile strength optimization

Abstract

Material-extrusion additive manufacturing produces polymer components with region-dependent mechanical behaviour arising from raster orientation, shell geometry, and layer-wise deposition. This study develops a region-based hybrid model for predicting the tensile strength of PLA specimens by separating the cross-section into top-and-bottom layers, perimeter shells, and an infill core. Response Surface Methodology was used to model the effects of layer thickness, print speed, and shell count within the relevant structural regions, while the Tsai–Hill criterion was used to represent orientation-dependent strength in the infill core. The segmented Design of Experiments reduced the calibration matrix from 252 to 72 configurations and the required specimens from 756 to 216, corresponding to a 71.4% reduction in experimental burden. Within the investigated range, tensile strength varied from 28.85 to 50.39 MPa. The standalone Tsai–Hill model produced errors of up to 11.52% at intermediate raster angles, whereas the calibrated hybrid model reduced the maximum error below 0.6%. Independent validation using 12 previously unused parameter combinations tested in triplicate produced a maximum absolute error of 0.69%. Further validation using non-standard specimen dimensions produced a maximum absolute error of 0.78%. Uncertainty propagation based on raw triplicate data produced a final propagated uncertainty of 0.144–0.159 MPa, corresponding to 0.38–0.42%, while a complete-specimen global RSM benchmark produced a maximum relative error of 5.13%, further supporting the benefit of the region-based decomposition strategy. Optical fracture-surface images supported the interpretation of raster-gap effects, although detailed polymer-interface phenomena were not directly measured. The framework is therefore presented as a validated engineering-scale predictive method for PLA within the investigated processing domain.