Dental diseases are among the most prevalent non-communicable conditions worldwide, with dental caries affecting billions of individuals and imposing substantial clinical and economic burdens in England and worldwide. These challenges are further exacerbated in medically compromised populations, particularly patients undergoing radiotherapy for head and neck cancer (HNC) related treatments, where irradiation significantly degrades the mechanical properties of enamel and dentine. This degradation increases susceptibility to fracture, secondary caries, and restoration failure, thereby reducing the longevity of conventional restorative materials. Hence, reliable dental restoration solutions should be explored. Although all-ceramic crowns are widely used due to their strength and aesthetic properties, their inherent brittleness and lack of hierarchical toughening mechanisms limit their performance, especially when bonded to structurally compromised dentine. This study presents a systematic experimental, computational, and data-driven investigation of nacre-inspired zirconia–polymethyl methacrylate (ZrO₂–PMMA) composites for load-bearing dental applications. Composites with varying zirconia volume fractions (65–80%) were fabricated using freeze casting and experimentally characterised to determine fracture toughness parameters (K_ic, K_jo, and K_jc). Finite element analysis (FEA) was developed and validated to simulate stress distribution, crack initiation, and propagation within nacre-like architectures. In parallel, machine learning (ML) models were implemented to predict fracture toughness and optimise material performance. The results identified the 75%-3-mol% yttria-stabilised tetragonal zirconia (3Y) combined with PMMA (polymethyl methacrylate) (3Y-75) composition as optimal, demonstrating superior fracture toughness and pronounced rising R-curve behaviour.

Furthermore, a physics-guided ML framework was developed to predict crack-path evolution using mechanistically relevant descriptors, including Young’s modulus (E) distribution, residual stress (RS), and ZrO₂ volume fraction (f). The framework accurately predicted crack trajectories and demonstrated the ability of deep learning to capture dominant fracture trends without explicitly solving governing fracture mechanics equations. This optimised composition was subsequently utilised in advanced three-dimensional FEA simulations of dental crowns cemented onto irradiated dentine, revealing a progressive reduction in load-bearing capacity with increasing radiation dose. Data-driven ML models accurately predicted both fracture behaviour and irradiation-induced mechanical degradation. In conclusion, the integrated experimental–computational–ML framework provides a robust and scalable approach for the design and optimisation of bioinspired dental restorative materials. The findings demonstrate the potential of nacre-inspired ZrO₂–PMMA composites to enhance fracture resistance, improve structural reliability, and enable personalised restorative solutions, particularly for structurally compromised and irradiated dentition.