【摘要】 目的为了有效地剔除测量数据中的噪声、提高结构损伤检测的诊断率.方法提出了一种新的消噪与结构损伤识别方法.采用小波/包变换对结构响应进行消噪处理,利用独立组分分析ICA中的固定点算法分离得到ICA滤波基并进行信号分离,接着对消噪、分离后的信号提取IC自相关特征,最后利用概率神经网络进行损伤识别.结果用Benchmark模型进行的损伤识别,并与进行二次小波/包消噪及未消噪的数据识别结果进行了比较结果验证了所提方法的可行性和有效性.结论笔者所提方法对信号消噪是可行、有效的,对提高损伤识别精度也有着重要的贡献.更多还原

【Abstract】 In order to effectively remove the noise of measured data and improve the diagnosis rate of a structure,a new approach of both de-noising and structural damage identification was proposed in this paper.Wavelet transform and wavelet packet transform were addressed according to the structural responses for denoising firstly,and then the fixed-point algorithm of the Independent Component Analysis(ICA) was used to process the denoised signal for acquiring the filtering base-function of ICA;afterwards,the signal was separated.The IC autocorrelation characteristics were extracted as soon as the signal was denoised and separated,and a prob-abilistic neural network(PNN) model was used for damage pattern recognition finally.To verify the method proposed,damage detection was conducted through the ASCE’s Benchmark model,and a comparison was made between the two proposed approaches,the one using twice denoising by wavelet and wavelet packet,and the one without any processing as well.The experimental study shows that the method proposed is effective for signal-noise separation,and provides crucial contribution to increase damage accuracy as well.更多还原