Enhancing Mechanical Performance of FFF-Fabricated PEEK Using an Integrated GA–ANN and FEA for Mandible Fracture Application

Polyether ether ketone (PEEK) is a radical filament with excellent strength equivalent to cortical bone and high thermal-mechanical properties. PEEK’s acquisition is acceptable in the fabrication of cranio-maxillofacial implants because of its exceptional strength-to-weight ratio and biocompatibility. However, its implementation in fused filament fabrication (FFF) is impeded by the lack of a cohesive optimisation framework that involves varying vital parameters: layer height, infill density and two post-process parameters: annealing temperature, annealing time, which affect its mechanical performance. This research work introduces a comprehensive methodology that integrates experimental design, hybrid Genetic Algorithm Artificial Neural Network (GA-ANN) modelling and finite element analysis (FEA) to improve the tensile properties of FFF fabricated PEEK. A total of 32 specimens (ASTM D638-IV) have been fabricated as per the central composite design (CCD) and tested for tensile strength using the Universal Testing Machine (UTM UNITEK 94100). The maximum tensile strength of 58 MPa is achieved at a layer height of 0.1 mm, 60% infill density and an annealing temperature of 250°C for a duration of 2 h. A hybrid GA-ANN is introduced to optimise process parameters, where ANN is trained using the Levenberg-Marquardt algorithm, achieving high predictive accuracy (R = 0.97375). GA-ANN has enhanced the tensile strength up to 3.34% at 0.13 mm layer height and 53.977% infill percentage at an annealing temperature of 202.47°C, with an annealing time of 2 h. A maximum von-Mises stress of 61 MPa is obtained through FEA validation. Moreover, a force-based case study on plate-screw assembly of mandible under physiological loading shows a maximum deformation of 0.92 mm, suggesting an adequate fixation stiffness supporting the mechanical suitability of the fixation construct. The experimental, numerical and GA-ANN-predicted results are in good agreement.