特徵擷取於橋樑非破壞檢測之應用

Application of Extracting Features on Bridge Nondestructive Testing

陳漢廷(Han-Ting Chen)

指導教授 ： 王安培

中原大學/工學院/土木工程研究所/碩士(2004年)

https://doi.org/10.6840/cycu200400412

摘要

本文為建立特徵值擷取應用於橋樑非破壞檢測的識別系統，此識別系統包含小波理論和模糊理論與類神經網路，可識別破壞程度及敲擊位置。
首先建立模擬橋面板模型進行實驗，模擬橋面板上鑽取各種不同的裂縫破壞程度，並於距裂縫不同距離的面板區塊上敲擊，藉由量測儀器以震波量測法記錄各種裂縫大小及損壞位置的振動訊號，再經由小波轉換作訊號處理，並以離散小波分解與重構擷取出各頻段訊號分布百分比作為擷取的特徵值，以此建立資料庫（database），接著將訊號的特徵值作為類神經網路輸入的特徵向量。可應用於識別橋面板不同裂縫大小及其敲擊位置。
本文分析步驟為使用小波分解重構擷取出特徵值再經由模糊理論作正規化再由機率類神經識別分析。由結果知，小波理論適用於訊號處理及特徵擷取，接著經由模糊理論作正規化可調整類神經輸入向量數目、並提供識別結果隸屬度參考，再經機率類神經訓練後，所建立的識別系統，可用於識別橋樑之破壞程度和敲擊位置，在本文假設條件下，經裂縫種類隸屬度檢核，識別幾可達100％正確。

關鍵字

機率類神經網路 ； 小波理論 ； 震波量測 ； 模糊理論

並列摘要

Abstract
II
Abstract
In this paper, an identification system is established to apply on
nondestructive testing（NDT）by extracting its features. This approach
comprises three phases applying wavelet theory, fuzzy logic, and neural
networks, respectively. It could distinguish not only destructive degree
but also breakpoint positions.
A crack with different size and breakpoint is experimented on the
bridge floor model, and the diverse vibrating signals are recorded by
seismic waves measuring method. Using the wavelet theory, the features
of the signals are extracted, and databases are established, which forms
the input eigenvector of ANN. The different crack sizes and breakpoint
positions can be judged through ANN’s training.
The features of the signals that extracted from discrete wavelet
decomposition and reconstruction are normalized using fuzzy theory
before probabilistic neural network（PNN）.
The results show that this approach can be applied to analyze signals
and extract the features appropriately. Additionally, the results also
show that inputs of PNN are adjusted through normalizing with fuzzy
theory, and it provides the degree of membership function. The
approach in this paper can be applied on NDT and identification of
breakpoint position. Accuracy of some PNN results may reach 100％.
Keywords：Wavelet theory, Fuzzy theory, Probabilistic neural network,
Seismic wave measure

並列關鍵字

Wavelet theory ； seismic wave measure ； Fuzzy theory ； Probabilistic Neural Network

參考文獻

1. Black, M., “Vagueness: an exercise in logical analysis”, Philosophy of science, Vol.4, pp.427~455,1937.

2. Daubechies, Ingrid “Orthonormal bases of compactly supported wavelets”, Communications on Pure and Applied Mathematics, Vol. XLI, pp.909-996, 1988.

4. Elkordy, M. F., Chang, K. C. and Lee, G. C., “Application of Neural Networks in Vibrational Signature Analysis”, Journal of Engineering Mechanics, Vol. 120, No. 2, February 1994.

6. Grossmann, A., and Morlet, J., “Decomposition of Hardy function into square integrable wavelets of constant shape”, SIAM J. Math. Anal. 15(4): pp. 736-783, 1984.

7. Hajmeer, M. and Basheer, I.“A probabilistic neural network approach for modeling and classification of bacterial growth/no-growth data”, Journal of Microbiological Methods, Volume: 51, Issue: 2, October, 2002, pp. 217-226