Research On Damage Identification Based On Quasi-static Displacement Influence Line Wavelet Transform And BP Neural Network

Bridge occupies an important position in transportation.Owing that the damage of bridge will cause huge economic loss,health monitoring is a topic worth studying.At present,There are still some defects in bridge damage identification method,such as the influence of environment,the difficulty of obtaining detection data and insufficient data analysis depth.Therefore,the quick and intelligent damage identification of bridge damage has become one of important research topics in bridge damage identification.In this paper,according to the existing research content,the bridge damage identification method based on quasi-static displacement influence line is studied as follows:(1)In view of problems that displacement time history curve under the action of moving load is greatly affected by the structural dynamic and static characteristics can’t fully reflect the overall information of bridge.In this paper,the influence line of static displacement is processed by continuous wavelet transform,and coefficients of wavelet transform are obtained.Then,according to the difference of wavelet transform coefficients before and after structural damage,the damage location index WTCD is established.At the same time,the area enclosed by curve of WTCD and the transverse axis is taken as the quantitative index WTCDA of damage.(2)The method of structural damage location is to locate damage location accurately by abscissa value corresponding to peak value of curve convex point of index WTCD.In addition,the quantitative method of structural damage is based on the data fusion ability of BP neural network.The WTCDA values of different damage degree and different measuring points at damage points are taken as the training samples of hidden layer.Then take WTCDA values of different measuring points at damage point to be measured as input layer data.Finally,the damage degree of the bridge structure is analyzed by the fusion of the data of different measuring points.(3)Through establishing the numerical models of single point and multi-point damage of simply supported and continuous box girders,the WTCD and WTCDA index are used to locate and quantify damage.Through numerical analysis of simply supported beam,load parameters and number of wavelet transform layers are determined,and the anti-noise analysis of WTCD and WTCDA index is carried out by using MATLAB to simulate Gaussian white noise.At last,arrangement and quantity of measuring points in the example of continuous beam are optimized.(4)In order to further verify the practicability of the damage identification method.In this paper,a quasi-static displacement influence line test is carried out on a simply supported steel plate beam.The effectiveness of measured data is improved by considering multiple groups of measured results under same test cast,combining S-G filtering method and data normalization method.Based on this expansion,single point and multi-point damage identification of simply supported beams is studied.The results show that: The damage location index WTCD can accurately locate single point and multi-point damage.WTCDA combined with BP neural network data fusion can accurately quantify degree of single point and multi-point damage.In addition,the damage identification method proposed in this paper can resist the noise with a signal-to-noise ratio of 30 d B,and has good noise resistance.Finally,the accuracy of the damage identification method is further verified by simply supported steel plate beam test.The quantitative accuracy of the test damage is reduced because of the uncontrollable factors in the test,but it is able to quantify the damage roughly.