Application of Machine Learning-Based Approach to Predict and Optimize Mechanical Properties of Additively Manufactured Polyether Ether Ketone Biopolymer Using Fused Deposition Modeling

Additive manufacturing (AM) or 3D printing has swiftly transitioned from a validation tool to a technology with substantial potential. Fused deposition modeling (FDM) is considered as one of the most widely used method for printing biopolymers such as polyether ether ketone (PEEK) which have immense potential application in various fields such as aerospace, biomedical, and automotive. This work aims to enhance this procedure by experimentally studying PEEK 3D Printing and employing machine learning approaches. Four printing parameters are investigated in this study, namely infill density, layer height, printing speed, and infill pattern that significantly influence mechanical strength, specific modulus, and surface roughness of the resultant prints. The topographical surface finish, ultimate tensile strength and elastic modulus of the printed parts are accurately predicted using ridge and Bayesian machine learning regression with permissible variance of less than 4% variation from actual measurements. The effects of process parameters on responses and hyperparameters on the model’s accuracy were studied using ‘Shapely value analysis and heatmaps,’ respectively. Furthermore, a genetic algorithm (GA) is utilized to optimize mechanical strength, resulting in a minimum surface roughness of 6.12 µm, a maximum ultimate tensile strength of 64.87 MPa, and an elastic modulus of 1267.30 MPa for an 80% infill density, 0.10 mm layer height, 25.10 mm/sec printing speed, and octet infill pattern. Parametric analysis and optimization results are supported by microstructure characterization and test results.