Structural damage localisation for a frame structure from changes in curvature of approximate entropy feature vectors

At present, accurate vibration-based damage localisation cannot be achieved very well in mechanical and civil engineering fields due to high noise in the measurements and low accuracy in finite element (FE) model of the measured structures. To address these issues, a method for damage detection is proposed in this work, i.e. the mean curvature difference method of approximate entropy (ApEn) feature vectors, based on the ApEn theory and curvature method. Simulation results of both single and multiple damage cases under pulse excitation indicate that the proposed method can be utilised to determine whether the damage is present in the structure or not and to locate the damage accurately, and the method exhibits strong anti-noise ability: it is feasible for damage with 5% stiffness reduction even if the noise level is up to 25%. Moreover, the proposed method does not require a structural FE model. Experimental results of a six-storey shear frame model also validated the proposed method. All of these lay a good foundation for its application in shear frame structures.