A novel damage evaluation method for timber utility poles based on wavelet packet transform and support vector machine

Publication Type:
Conference Proceeding
SHMII 2015 - 7th International Conference on Structural Health Monitoring of Intelligent Infrastructure, 2015
Issue Date:
Full metadata record
© 2015, International Society for Structural Health Monitoring of Intelligent Infrastructure, ISHMII. All rights reserved. This paper presents a novel damage evaluation approach using wavelet packet transform (WPT) and support vector machine (SVM) for the damage assessment of timber utility poles. The method is based on guided wave (GW) testing using a multi-sensor system. In the new method, first, WPT is utilized to extract entropy features of GW signals. Then, to eliminate the multicollinearity between extracted features, principle component analysis (PCA) is adopted and entropy features are replaced by a few principle components. Finally, a classifier model based on SVM is constructed to assess the pole condition. To improve the estimation accuracy of the model, genetic algorithm (GA) is used to optimize the parameters in the SVM. The new method is validated on laboratory timber specimens (undamaged and damaged) that are experimentally tested using an impact hammer for wave excitation and a multi-sensor array for wave response recording. WPT-based entropy feature extraction and PCA is subsequently applied to the recorded wave signals and the damage condition of the timber specimen is identified using the pre-trained classifier. The experimental results verify that the proposed method is effective achieving a high identification accuracy of up to 95%.
Please use this identifier to cite or link to this item: