Numerical simulation of soft stone masonry under simple compression using the applied element method

Soft stone masonry, such as tuff stone masonry, characterizes many buildings in southern Italy, usually in the form of vaults and load-bearing walls made up of regular, single-leaf or multi-leaf, assemblages with mortar joints. Widespread use of such a masonry type in historical non-engineered buildings calls for numerical tools able to accurately simulate the behaviour of soft stone masonry against various kinds of load. As shown by past experimental tests and in-situ surveys, numerical simulation should account for complex failure patterns involving both joints and units. Particular care should be paid to local phenomena such as splitting and crushing failures, which are often seen, in post-earthquake scenarios, as a result of flexural and rocking failures of masonry piers. Simplified micro-modelling approaches, which discretize masonry through assemblies of (i) continuum media representative of masonry units and joints or (ii) discontinuous sets of bodies connected by zero-thickness interfaces, might yield such advanced output. An advantage of these latter strategies over the former, when implemented in a discrete or applied element method framework, is that material fracture, separation and collision can be explicitly simulated. However, a common assumption of these methods is to use rigid bodies, the suitability of which to model soft stone behaviour calls for in-depth experimental validation, more so in the case of compressive loading and related failure mechanisms, which play a significant role in the response of masonry structural elements towards both vertical and horizontal loading conditions. In this paper, the suitability of simplified micro-modelling based on the applied element method (AEM) for numerical simulation of soft stone masonry under compressive loads is investigated. Three different AEM-based numerical modelling approaches are validated against comprehensive sets of experimental data, accounting for loading conditions perpendicular and parallel to bed joint orientation. Their accuracy is then discussed in terms of failure patterns and load-displacement curves, with a focus on the issue of output quality against computational time and required input data. Special attention is paid to the modelling of masonry crushing, producing a significant step forward in AEM simulation and realistic reproduction of masonry behaviour under severe loading scenarios.