Soil settlements are one of the causes of damage to historic masonry buildings, together with mechanical properties deterioration and environmental actions. The size of the settlements is often unknown, as most often measurements in time are not available. The crack pattern provides information on the members where settlements occur, and monitoring can measure the rate in time. The aim of the paper is structural analysis for the assessment of the piers and foundations and decision making for strengthening works. A case study of the arched bays on the interior walls of a historic church is chosen. A limit analysis model with rigid blocks and interfaces is set up; this includes as variables the masonry strength properties for shear friction and cohesion. Numerical crack predictions are compared to crack pattern surveys, to find the material properties with the effects closest to the observations. The results of the model are then compared to a nonlinear finite element model, as a verification of the results of the simplified rigid block analysis. The limit analysis model is used in addition to generate results on pillar reactions in relation to different settlement patterns. Finally, a machine learning algorithm is used to establish a relation amongst these quantities, to be used in the design of foundation strengthening.