Optimization of mechanical performance in bioinspired FDM-printed PLA sandwich structures using machine learning

Despite some successes and claims of the best possible mechanical performance in 3D printed parts, it is essential that intelligent design approaches be linked to intelligent process control. In this paper, an integrated experimental-machine learning methodology is presented to optimize the polylactic acid (PLA) sandwich structures that has bio-inspired internal cores. They are made using Fused deposition modeling (FDM). Six basic geometries AS, ASR, ASVC, HAR, HAS and HASV were printed with three orientations 0 0 , 45 0 and 90 0 to have 54 specimens. Mechanical properties, including maximum load, ultimate tensile strength, Young’s modulus, tensile stress at break, and tensile strain, were measured and studied with the help of the Taguchi method and ANOVA, which held core structure and raster orientation as having significant effect. The HASV core with 0 0 raster orientation reached the maximum load value of 1426.3 N, maximum UTS value of 18.3 MPa, and Young’s modulus value of 933.7 MPa, which is 168% stronger than the weakest configuration (HAS, 45 0 , 532 N). ASR at 0° exhibited the greatest ductility (tensile strain 12.79%). The ANN model with the usage of Bayesian Regularization (10 neurons) provided a training R = 0.975 and the mean 5-fold-cross validation R = 0.962, proving a good predictive ability of the model without overfitting. These results offer an effective way forward to the design of PLA parts facing high loads, proving that statistical optimization, combined with AI, is the way forward in the field of additive manufacturing.