ann

以下为《ann》的无排版文字预览，完整内容请下载

Engineering Structures 52 (2013) 408–421 Contents lists available at SciVerse ScienceDirect Engineering Structures journal homepage: doc.001pp.com/locate/engstruct The application of a damage detection method using Arti?cial Neural Network and train-induced vibrations on a simpli?ed railway bridge model Jiangpeng Shu ⇑, Ziye Zhang, Ignacio Gonzalez, Raid Karoumi Department of Civil and Architectural Engineering, KTH Royal Institute of Technology, Stockholm, Sweden article info Article history: Received 15 May 2012 Revised 4 December 2012 Accepted 25 February 2013 Available online 10 April 2013 Keywords: Damage detection Railway bridge Dynamic response Statistical property Arti?cial Neural Network (ANN) abstract A damage detection algorithm based on Arti?cial Neural Network (ANN) was implemented in this study using the statistical properties of structural dynamic responses as input for the ANN. Sensitivity analysis is performed to study the feasibility of using the changes of variances and covariances of the dynamic responses of the structure as input to the ANN. A ?nite element (FE) model of a one-span simply supported beam railway bridge was developed in ABAQUSÒ, considering both single damage case and multi-damage case. A Back-Propagation Neural Network (BPNN) was built and trained to perform damage detection. A series of numerical tests on the FE model with different vehicle properties was conducted to prove the validity and ef?ciency of the proposed approach. The results reveal not only that the ANN, together with the statistics, can correctly estimate the location and severity of damage, but also that the identi?cation of the damage location is more dif?cult than that of the damage severity. In summary, it is concluded that the use of statistical property of the structural dynamic responses as damage index along with the Arti?cial Neural Network as tool for damage detection for an idealized model of a bridge is reliable and effective. Ó 2013 Elsevier Ltd. All rights reserved. 1. Introduction In attempt to determine the occurrence, location and the severity of any damage to structures, various detection methods based on dynamic response have aroused considerable interest to civil engineers. In the past few decades, intensive research have been devoted to the Arti?cial Neural Network (ANN) aided damage identi?cation technique, a typical approach when the modal features are not well known. ANN is a mathematical model capable of implicitly detecting complex nonlinear relationships between dependent and independent variables with the ability to self-learn and fault-tolerance. These bene?ts make it appropriate for minimizing the negative impacts of uncertainty in responses measurements and FE model con?guration [1]. Yeung and Smith [2] used unsupervised neural network for pattern recognition with the data ?ow generated by the instruments installed on Tsing Ma Bridge to continuously examine structural performances. It was shown that the neural networks could be adjusted to permit damage detection even in the presence of noise in the signals. Given that damage did not occur as a Boolean relation (one of two values, true or false) but progressively, Reda Taha and Lucero [3] introduced a method by supplementing ANN aided ⇑ Corresponding author. Tel.: +46 ***8. E-mail address: jiangpeng.shu@chalmers.se (J. Shu). 0141-0296/$ - see front matter Ó 2013 Elsevier Ltd. All rights reserved. http://dx.doi.org/10.1016/j.engstruct.2013.02.031 identi?cation with fuzzy set to accommodate uncertainty associated with ambiguous damage states. Accelerations from sensors distributed over the bridge were analyzed using a wavelet-neural network module to establish patterns of dynamic behavior of the bridge. They examined the algorithm using data simulated from ?nite element analyses of a pre-stressed concrete bridge without a prior knowledge of damage level and proved it capable of identifying damage accurately. Li [4] introduced probabilistic neural networks to process dynamic signals from a data collection system. He de?ned the damage cases into 16 categories by grouping neighboring elements to facilitate the Probabilistic neural networks easy implementation. The limitation of Li’s study is that the number of categories may not include all types of structural damage and only single damages can be detected. Nyarko et al. [5] implemented a multilayer perception (MLP) neural network to model the relationship between the structure parameters (natural period, elastic base shear capacity, post-elastic stiffness and damping) of an SDOF model and the damage ratio (DR) coef?cient. A new original formula for damage ratio coef?cient was employed for performing sensitivity analyses on the trained MLP neural network to exam the damage level of a bridge after earthquake. Li and Yang [6] developed a technique using Arti?cial Neural Network based on statistical properties of structural dynamic response exited from white noise. Back-propagation ANN with the changes of variance of structural response as input vector and damage status as output J. Shu et al. / Engineering Structures 52 (2013) 408–421 409 Nomenclature A area of cross section of beam c coef?cient of viscosity of single-degree-of-freedom sys- tem C matrix of coef?cient of viscosity of multiple degree-of- freedom system Cov(t1,t2) covariance of outputs, t1 and t2 are discrete time points d bogie axle spacing D coach length E Young’s modulus E[x(t)] expectation of outputs ep ððEEIIÞÞdiui n f(t) ratio of standard deviation value between noise and signal stiffness of structure after damage stiffness of structure before damage external loads with respect to single degree-of-freedom system fs covariance relationship function of single degree-of- freedom system fm covariance relationship function of multiple degree-of- freedom system F(t) external loads with respect to multiple degree-of-free- dom system F(x) Fourier transformation of f(t) FÃðxÞ conjugate function of F(x) h number of neurons in the input layer h(t) unit-impulse response H(x) transfer function I moment of inertia k stiffness of single-degree-of-freedom system K matrix of stiffness of multiple degree-of-freedom sys- tem Kkj K Ã ln L m M n N N0 P1,P2 r Rx(t1,t2) s Ssjn x(t) X ym yc ai x0 x1, x2 q t n r(yc) D /ki, /ji function of modal shape (/ki, /ji) and transfer function (Hjn) conjugate function of Kln span length of bridge beam mass of single degree-of-freedom system matrix of mass of multiple degree-of-freedom system total number of elements number of degree of freedom standard normal distribution axle load number of neurons in the hidden layer autocorrelation function of outputs, t1 and t2 are discrete time points time impulse power spectra density function displacement responses of single degree-of-freedom system matrix of displacement responses of multiple degree-offreedom system polluted response intact response stiffness reduction of i-th element nature circular frequency of structure circular frequency with respect to time domain density Poisson’s ratio damping ratio standard deviation of unpolluted response relative distance between the detected element and the pre-de?ned element components of mode-shape matrix U is adopted for identifying the damage in a beam. The numerical results show that the proposed method can accurately detect damage location and extent. It remains to be studied how the applicability of the proposed method is affected by the variation of the system properties such as the location of damage, measuring noise, vehicle load and speed. Therefore the present study was carried out to sequentially investigate ef?ciency of back propagation neural network (BPNN) utilized in damage detection for a bridge based on statistical properties of dynamic response. Live loads were adopted instead of white noise excitation, because of it superior noise resistance. The main objectiv 内容过长，仅展示头部和尾部部分文字预览，全文请查看图片预览。 itivity and its application in damage detection. J Sound Vib 2007;303(1-2):305–29. [13] Kesavan A. Embedded intelligence in structural health monitoring using arti?cial neural networks, PhD thesis. School of Aerospace, Mechanical and Manufacturing Engineering, RMIT University, Australia; 2006. [14] Kolmogorov AN. On the representation of continuous functions of many variables by superposition of continuous functions of one variable and addition. Dokl Akad Nauk SSSR 1957;114:953–6. in Russian. [文章尾部最后500字内容到此结束,中间部分内容请查看底下的图片预览]请点击下方选择您需要的文档下载。

以上为《ann》的无排版文字预览，完整内容请下载

ann由用户“wj9617”分享发布，转载请注明出处

XXXXX相关资讯

XXXXX猜你喜欢

回顶部 | 首页 | 电脑版 | 举报反馈更新时间2020-03-13 12:26:08