Damage detection in a cantilever beam using noisy mode shapes with an application of artificial neural network-based improved mode shape curvature technique

Experimentally and numerically obtained displacement mode shapes are utilized as input data to artificial neural networks (ANNs) and mode shape curvature technique. Frequency responses (FRs) in the form of displacement mode shapes with varying damage levels are extracted using the Bruel & Kjaer instrument. Two identical specimens of a cantilever beam are considered with different damage locations. It is demonstrated that the measured frequency response needs to be made error-free to locate damages. ANN training algorithms are utilized to reduce the measurement error from the measured frequency response (FR) data set. The analysis is more robust due to the use of ANN application before extracting mode shape curvature. The trained data sets are then utilized to produce the mode shapes curvatures for all the damage cases using central difference approximation. Damage severity and locations are then identified by analyzing the absolute mode shape curvature differences in different damage scenarios. © 2021, The Author(s), under exclusive licence to Springer Nature Switzerland AG.