A better experimental method for damage detection of composite plate using principal component analysis (PCA) and nearest neighbor principle

Aim. The introduction of the full paper reviews Refs.3 through 7 and points out what we believe to be their respective deficiencies; then, we propose what we believe to be a better method. We explain our method in sections 1, 2 and 3, whose core consists of: (1) we used the frequency response functions (FRFs) of the intact and damaged plates as input data and measured the FRFs of specific nodes of the composite plates by sweeping sinusoidal excitation; (2) we applied the PCA method to reducing the number of dimensions of the FRFs and used, instead of raw FRFs data, as training sample sets in the nearest neighbor principle, the first 13 orders of their principal components which contain the majority of structural health information; (3) we used the nearest neighbor principle to recognize the damage states of the tested FRFs of the composite plates. Section 4 presents the damage detection results of three training sample sets; the detection results, given respectivedly in Tables 2, 3 and 4, and their analysis show preliminarily that our damage detection method can correctly and effectively detect the damage state of a composite plate.