Review on Non-Destructive Prediction of Recycled Aggregate Crushing Value

The resource utilization of recycled aggregates is of paramount importance, and the crushing value serves as a core indicator for evaluating their mechanical properties and classification applications. This paper systematically reviews the research progress on the non-destructive prediction of the crushing value of recycled aggregates and classifies existing approaches according to their primary information sources into physics-based DEM simulation, physical-indicator-based data-driven prediction, and image-based deep-learning prediction. The findings indicate that DEM has been widely used to reveal the mesoscopic crushing mechanisms of particles, but studies that directly output the standard crushing value of recycled aggregates remain limited and are often semi-quantitative or validation-oriented. Although physical-indicator-based data-driven methods and image-based deep-learning methods can establish nonlinear relationships between physical indicators or morphological features and the crushing value, neither approach can directly perceive internal defects within aggregates, and both may exhibit limited generalization capability. Based on these findings, this paper envisions a multi-modal synergistic prediction direction centered on the deep integration of “physics and data.” By incorporating mechanical prior constraints and multi-source information fusion mechanisms, this approach is expected to facilitate the paradigm shift in crushing value detection from “destructive post hoc validation” to “intelligent real-time prediction.”