Neutrosophy and AI – Advances in SHM: Status, Location, Severity, Causality, and Improvements

Structural health monitoring (SHM) must detect damage, locate it, estimate its severity and identify its cause. This paper combines CAX (Computer-Aided Extended) tools, artificial intelligence (AI) and neutrosophic methods for this purpose. A parametric finite element model of a three-level structure (353,544 hexahedral elements) was used to generate automatically 4,536 models with different damage locations, severities and causes (mechanical stiffness reduction or thermal load), which were used to train an artificial neural network (ANN). In the cases presented, the trained network located single and multiple damages and distinguished mechanical from thermal causes; for example, a 33% stiffness reduction in column C31 at level 2 was assigned to the nearest severity class in the database (25%), and a 150 °C thermal load to the 160 °C class. Neutrosophic (T, I, F) triplets computed from the modal response are used for damage localization and as ANN inputs, several neutrosophic score functions are discussed, and a neutrosophic score is compared with the modal assurance criterion (MAC) for the pairing of eigenmodes. A post-detection procedure that adjusts the stiffness of the diagonal elements brought the resonant frequency of the damaged structure back close to that of the intact one.