A wavelet analysis-based approach for damage localization in wood beams

Abstract

Free vibration testing was conducted to generate the first two mode shapes for damage detection in timbers. A wavelet transform was proposed to postprocess the mode shapes for damage pattern recognition. The wavelet used here was "db3." The different damage severities, damage locations, and number of damaged areas were simulated by removing mass from intact beams. The results showed that the chosen wavelet db3 is suitable and that the wavelet coefficients are sufficiently sensitive to identify the existence of damage and its location in cases of different damage location, severity, and number. An edge distortion effect was apparent at the two computing edges where the wavelet coefficients were abnormally high. The wavelet coefficients showed dominant spikes around the damage locations and were zero everywhere else except the two computing edges. The dominant spikes coincided well with the damage location. © 2006 The Japan Wood Research Society.