Application of Improved Distance Evaluation Technique in feature selection for crack detection of a steel beam

In this paper a vibration experiment for cantilever beam with a single-edge crack is implemented in order to observe the most distinguished changes in vibration characteristics, including modal parameters and extracted statistical features from signals. Hence, an experimental setup is designed in which the cantilever beam is excited by modal hammer and shaker while the acceleration is measured by the accelerometer mounted at the beam surface. At signal processing level, Fast Fourier Transform and Wavelet Transform are utilized. To assess the changes of beam characteristics in physical aspects, experimental modal analysis is implemented while for understanding signal features some statistical features are extracted from time, frequency domain response, such as mean, minimum, and maximum, kurtosis, etc. Moreover, the similar features are extracted from Wavelet coefficients. These features finally are compared with the healthy beam to acquire the most distinguished features. A total set of 20 features are extracted from each signal domain. Then, Improved Distance Evaluation Technique (IDET) is used in order to select the most superior features for crack detection of the steel beam. These superior features are selected based on the ability to distinguish the beam conditions. The results of this research show that the loss of natural frequencies in the 1st and 2nd modes is among the distinguished features since their sensitivity to the crack is reasonably high. Applying IDET to the different features strongly clarifies the capacity of time and Wavelet extracted features over the FRF features, which conclusively makes the sensitivity of time and Wavelet features to the crack distinct.

کلیدواژه ها:

Cracked beam ، Modal analysis. FFT ، Wavelet transform ، Feature extraction ، Superior features ، Improved Distance Evaluation