سال انتشار: ۱۳۸۶

محل انتشار: پنجمین کنفرانس بین المللی زلزله شناسی و مهندسی زلزله

تعداد صفحات: ۸

نویسنده(ها):

Qiao – Graduate Research Assistant, Dept. of Civil Engineering, Kansas State University, Manhattan, KS 66506-5000, USA
Esmaeily – Assistant Professor, Dept. of Civil Engineering, Kansas State University, Manhattan, KS 66506-5000, USA
Melhem – Professor, Dept. of Civil Engineering, Kansas State University, Manhattan, KS 66506-5000, USA,m

چکیده:

Structures will be pushed into their non-linear response and experience severe damage under large seismic forces. A signal-based pattern-matching procedure is used for structural damage detection with a limited number of input/output signals. The key of the method is to extract sensitive features of the structural response signals in the form of patterns, which present their unique conditions under different damage scenarios, and select efficient pattern recognition algorithms that best perform pattern matching and classification. Two types of transformation algorithms for feature extraction were investigated separately when processing the signal: (1) Fast Fourier Transform (FFT) to extract frequency- based features, i.e. FFT magnitudes that form a one-dimensional pattern; and (2) Continuous Wavelet Transform (CWT) to extract time-frequency-based features, i.e. CWT coefficients that form a twodimensional pattern. Three pattern recognition algorithms were further investigated individually to perform pattern-matching: correlation, least square distance, and Cosh spectral distance. A damage pattern database was developed analytically for a wide range of damage severities and locations. Damage location and severity were identified by best matching the feature of the response signal (after an unknown damage) against the developed database. For demonstration purposes, some cases were numerically simulated and studied for a three-story steel structure. The results show that different damage scenarios can be uniquely presented by these transformations, and correlation algorithm can best perform pattern matching to identify the damage location and severity correctly evenwhen the signal iscontaminated with noises. The proposed method is suitable for structural health monitoring, especially for online monitoring applications. It can potentially be utilized in detection of the type of damage in addition to the location and severity.