Determination of Composition of Masonry Mortars for Conservation of Historical Constructions Using Artificial Neural Networks.

This study presents a novel approach to determine the composition of masonry mortars and their types from cement, lime, and cement-lime using an artificial neural network (ANN). It also allows the preparation of mortar recipes for the conservation of historical masonry objects with properties similar to the original ones, but using currently available raw materials. An ANN was trained using a set of cement, lime, and cement-lime mortars with known compositions. The properties chosen for the ANN's analysis included total porosity, specific density, insoluble residue content, silicone (SiO) content, calcium (CaO) content, Si/Ca ratio in grout, and compressive strength. The use of ANNs allows for the determination of mortar composition with a validation error of less than 5% and a method of classification of the type of mortar that gives correct answers in more than 93% of cases, proving the usefulness of ANNs in determining the type and composition of masonry mortars relevant for the conservation of historical masonry structures.