A novel method of vibration modes selection for improving accuracy of frequency-based damage detection

Frequency-based damage detection techniques have been widely applied to structural health monitoring. By analysing the changes (shifts) in natural frequencies in a structure with and without damage, these techniques solve the inverse problem of determining size and location of damage. In the existing literature, the first few or random modes of frequency shifts are given to the inverse algorithms as inputs in order to predict the damage parameters. These frequency shifts can be either numerical or measured. While the accuracy of prediction in the former (numerical) case has been found to be satisfactory, the use of measured frequencies has often shown large errors. This can be attributed to unavoidable noise in frequencies, including the mismatch between FEM model and real structure, as well as the noise in the measurement itself. Previous research has shown that the noise in frequency will actually be magnified in the discrepancy of frequency shifts, and thus affect the damage prediction accuracy. And moreover, the same noise added to different modes of frequency of a damaged case will lead to the different levels of deviation in different modes of frequency shifts. This observation indicates that potentially some modes of frequency shifts are less affected by the noise than others for a given case. However, so far, there has been no studies that attempt to identify particular vibration modes of frequency shifts that are (a) less affected by the noise for all damage cases and (b) result in a more accurate prediction of damage. In this study, a novel concept of Noise Response Rate (NRR) is proposed to evaluate the sensitivity of each mode of the frequency shift to noise. Further, it is shown that selecting the vibration modes with low NRR values improves the prediction accuracy of frequency-based damage detection. The efficacy of NRR is demonstrated through a case study on a composite curved plate compared with the conventional method for damage detection.