Guided-wave-based damage detection of timber poles using a hierarchical data fusion algorithm

Publisher:
Southern Cross University
Publication Type:
Conference Proceeding
Citation:
Proceedings fo the 23rd Australasian Conference on the Mechanics of Structures and Materials, 2014, pp. 1203 - 1208
Issue Date:
2014-12-09
Full metadata record
Files in This Item:
Filename Description Size
ACMSM23-147-CameraReady.pdfAccepted Manuscript version530.06 kB
Adobe PDF
This paper presents a hierarchical data fusion algorithm based on the combination of wavelet transform (WT), back propagation neural network (BPNN) and Dempster-Shafer (D-S) evidence theory for the multi-sensor guided-wave-based (GW-based) damage detection of in-situ timber utility poles. In the data-level fusion, noise elimination is performed on the original wave data to obtain single-mode signals using WT technology. Statistical information is extracted from the single-model signals as major characteristic parameters. In the feature-level fusion, for each sensor in the testing system, two sub-networks corresponding to different types of GW signals are constructed based on BPNN and characteristic parameters are sent to the networks for initial state recognition. In the decision-level fusion, the D-S evidence theory method is adopted to combine the initial results from different sensors for final decision making. The overall algorithm employs a hierarchical configuration, in which the results from the former level are regarded as input to the next level. To validate the proposed method, it was tested on GW signals from in-situ timber poles. The obtained damage detection results clearly demonstrate the effectiveness and accuracy of the proposed algorithm.
Please use this identifier to cite or link to this item: