Determination of piers shear capacity using numerical analysis and machine learning for generalization to masonry large scale walls

Abstract

In this study, comprehensive investigations were conducted on the structural behavior of masonry shear walls considering both aspect ratios and the axial forces. A reliable and simplified macro approach called Multi Pier (MP) method was employed to analyze single masonry piers with different aspect ratios (h/l). The shear and flexural strength along with the failure mechanism of each pier were obtained versus various levels of axial force during the sensitivity analysis. The results of the analysis were used as an input of a machine learning approach called Back Propagation Multi-Layer Perceptron (BPMLP) to generalize the data. The outcomes of machine learning was utilized to develop simple, but accurate equations as well as a user-friendly application to obtain the shear strength of a masonry pier according to the mechanical and geometrical properties. The BPMLP network is also able to produce and estimate all potential failure modes of a masonry wall. Finally, the efficiency of the proposed equations and developed application was evaluated using a novel macro-element approach. It is concluded that the proposed equations can be used in any macro-elements approach to enhancing efficiency and saving time. The proposed Equivalent Wall (EW) model was able to precisely estimate the response of full-scale façade masonry walls subjected to the in-plane loads. According to this model, each pier and spandrel was substituted by two elastic continuous (2D) macro-elements which connected via two bi-component discrete elements. A linear behavior based on masonry modulus of elasticity and bi-linear constitution according to the shear strength was assigned to the axial and shears components of the discrete links, respectively. The proposed EW model was verified using a five-story-seven-bay masonry wall. Satisfactory results were observed in terms of global load-deflection behavior comparing with popular approaches from the literature.