Mortarless interlocking brick structures offer great potential for modern masonry construction due to their advantages in construction efficiency, cost-effectiveness, and environmental sustainability. However, their complex geometry and highly nonlinear contact interactions present significant challenges for accurate numerical simulation. To address these challenges, this study proposes an efficient computational framework that integrates the rigid-body-spring model (RBSM) with a machine-learning-based parameter identification and optimization strategy. A homogenized model that captures the geometric characteristics of a novel interlocking brick unit is developed to analyze the dynamic response of large masonry walls subjected to in-plane cyclic loading. In this framework, the multi-linear constitutive models of the interfacial springs in the RBSM are derived and simplified from the tensile–compressive behavior of concrete, with initial control-point parameters determined based on the stiffness of contact surfaces. Using these initial parameters, sample datasets are generated via Latin hypercube sampling, followed by batch dynamic simulations employing both the RBSM and finite element (FE) models. An artificial neural network (ANN) is trained to predict the error between the two models, and the RBSM parameters are subsequently optimized. Finally, the optimized RBSM is validated against experimental results of large-scale masonry walls, demonstrating excellent agreement. The proposed numerical approach effectively overcomes key limitations in the analysis of interlocking masonry systems, providing a reliable and computationally efficient tool for predicting the dynamic performance of similar structures. • A machine-learning-assisted RBSM framework enables efficient cyclic analysis of interlocking masonry. • Machine learning automates optimization of dynamic constitutive parameters in RBSM. • The optimized model accurately predicts cyclic response of full-scale interlocking walls. • A unique three-stage failure mechanism of interlocking systems is revealed.
Lu et al. (Mon,) studied this question.