Structural Performance of Crumb Rubber-Modified Asphalt Under Service Conditions: A Mechanistic and Machine-Learning Assessment

Crumb rubber (CR) modification offers a sustainable approach to improving asphalt performance while promoting the beneficial reuse of end-of-life tyres. This study evaluates the structural performance of CR-modified hot mix asphalt (HMA) under service conditions through an integrated mechanistic and machine-learning framework. Laboratory-measured resilient moduli for conventional and 15% CR-modified HMA at 20, 40, and 60 °C were incorporated into a three-layer Odemark–Shell analysis considering 11 asphalt-layer thicknesses and three representative seasonal temperature periods established from an 11-year (2015–2025) climatic record for Karbala, Iraq. Pavement performance was assessed for a 20-year design period under 6 × 106 equivalent single axle loads (ESALs), with fatigue and rutting damage accumulated using Miner’s rule. Eight machine-learning algorithms were additionally evaluated as fast surrogate models for predicting the critical pavement responses. Incorporating CR increased the resilient modulus by 8.5–9.2%, resulting in reductions of 9.7–10.4% in horizontal tensile strain and 1.2–2.8% in vertical compressive strain over the investigated temperature range. The continuous interpolation of the fatigue-damage criterion indicated required asphalt thicknesses of approximately 27.3 cm for conventional HMA and 25.8 cm for CR-modified HMA, corresponding to a reduction of approximately 1.5 cm (5–6%). At the common 27.5 cm design thickness, CR modification reduced the governing fatigue damage index from 0.954 to 0.723, representing a 24.2% reduction in accumulated fatigue damage and increasing the estimated service life from approximately 21.0 to 27.7 years. Gradient Boosting achieved surrogate-model fidelity, with R2 values of 0.993 and 0.9998 for the critical strain responses, while asphalt-layer thickness emerged as the dominant performance variable, consistent with the known analytical structure of the governing equations. The use of CR could additionally divert approximately 10.9 tonnes of scrap tyre rubber from the landfill per lane-kilometre. The findings demonstrate that the CR modification can reduce the continuously estimated required asphalt thickness and extend the theoretical pavement service life under the adopted assumptions, although the thickness reduction is not resolved when standard 2.5 cm construction increments are considered. Additionally, CR modification provides measurable sustainability benefits through the reuse of scrap-tyre rubber.