Artificial Neural Network Model for Analysis of In-Plane Shear Strength of Partially Grouted Masonry Shear Walls

Abstract

The behaviour of partially grouted (PG) masonry shear walls is complex, due to the inherent anisotropic properties of masonry materials and nonlinear interactions between the mortar, blocks, grouted cells, ungrouted cells, and reinforcing steel. Since PG shear walls are often part of lateral force resisting systems in masonry structures, it is crucial that its shear behaviour is well understood, and its shear strength is accurately predicted. This study presents the development of an artificial neural network (ANN) model for analyzing the shear strength of PG walls. ANNs have the unique ability to address highly complex problems and the potential to predict accurate results without a defined algorithmic solution. By providing an ANN with a dataset of multiple inputs and corresponding outputs, it can be trained to describe nonlinear relationships that may exist among the variables and provide insight into the influence of each input parameter. An experimental dataset of PG shear walls is used as input for the ANN analysis model. It is necessary to assemble the dataset from multiple experimental studies using meta-analysis, given that no single experimental study contains enough information to build and validate a constitutive model for the shear strength and behaviour of PG walls. Finite element (FE) modelling is shown to be a viable option for addressing gaps in input values which exist in the dataset. The effect of previously unaccounted parameters in code-based approaches is discussed, as well as the influence of different types of ANN analysis options and input size on the model predictions. The ANN model results are compared against currently available design codes and equations to predict the in-plane shear strength of PG shear walls.