Intelligent identification of crack modes and damage evolution in SCA-modified metakaolin-based geopolymers using AE, DIC, and CNN

Silane coupling agents (SCAs) are widely recognized for improving the mechanical performance and crack resistance of geopolymers. However, previous studies have primarily focused on strength enhancement, while the influence of SCAs on crack mode evolution and micro-damage development remains poorly quantified. This study investigates the fracture regulation and damage evolution of methyltrimethoxysilane (MTMS)-modified metakaolin-based geopolymers through an integrated analysis of mechanical behavior, acoustic emission (AE), digital image correlation (DIC), and convolutional neural networks (CNNs). MTMS was incorporated at dosages ranging from 0 to 0.4 wt%, and CNN-based crack mode identification model was developed using a balanced AE dataset containing 1200 tensile and shear cracking signals. The results show that MTMS regulates both strength and fracture behavior in a dosage-dependent manner. The optimum strength enhancement was achieved at 0.2 wt% MTMS, increasing compressive and flexural strengths by 22.9% and 30.8%, respectively, while 0.3 wt% MTMS produced the highest peak displacement and longest time to peak load. AE analysis revealed a transition from instantaneous brittle damage to progressive energy-dissipating fracture with increasing MTMS dosage, whereas excessive MTMS induced defect-related brittleness recovery. The 2D-CNN achieved a validation accuracy of 96.67%, enabling reliable quantification of tensile and shear cracking evolution. CNN classification showed that 0.2 wt% MTMS promoted mixed-mode fracture by reducing tensile crack dominance and increasing shear-related damage activity. Combined DIC-AE analysis identified three fracture stages, including microcrack initiation, stable crack propagation, and unstable fracture. This study provides an intelligent framework for quantitatively linking SCA-induced fracture regulation with micro-damage evolution in geopolymers.