Modeling of masonry failure surface under biaxial compressive stress using Neural Networks

• Neural Networks are used in order to approximate the experimental results for masonry failure.
• A two-step procedure is proposed, with the training of two types of Neural Networks.
• The NNs showed great performance in fitting the experimental input data.
• The curves generated by the NNs are continuous and smooth, but not necessarily convex.

Masonry is a brittle anisotropic material that exhibits distinct directional properties because the mortar joints act as planes of weakness. To define failure under biaxial stress, a 3D surface in terms of the two principal stresses and their orientation to the bed joints, is required. In the present study, a novel method is proposed on applying Neural Networks (NNs) to approximate the failure surface for such brittle anisotropic materials. The method comprises a series of NNs that are trained with available experimental data. The results demonstrate the great potential of using NNs for the approximation of masonry failure surface under biaxial compressive stress.