摘要
用小波和神经网络ART2相结合的方法检测结构的损伤位置。给出了小波变换和人工神经网络的基本理论及其用于损伤检测的原理与特点。通过把小波变换作为神经网络的前处理来构造小波神经网络。首先通过数值试验检验了小波消噪和小波神经网络损伤检测的能力。然后在一个框架结构模型上进行了试验。实验证明这种方法使网络抗噪声能力增强,使损伤识别的效果更好。ART2网络具有自动从环境中学习的能力,能自动的给出新的识别输出。
The application of wavelet-based neural network ART2 to the damage detection of structure is discussed. A method combining dyadic wavelet with neural network of ART2 is presented and the damage location can be well identified with this method. The basic theories of artificial neural network and wavelet transform are given and their features and the principle of damage detection are analyzed. Wavelet-based neural network is constructed by taking wavelet transform as the pre-processor of neural network. Then the wavelet de-noise, the detection of changes of a signal and the ability of damage detection of wavelet-based neural network are tested by numerical samples. The effectiveness of this method is attested further by a model frame structure. The results show that the present method is feasible and it has advantages of few requirements of historical data, automatic increase of identification category, and the noiseproof ability.
出处
《工程力学》
EI
CSCD
北大核心
2003年第6期176-181,共6页
Engineering Mechanics
