Research On Bridge Damage Identification Based On Modal Entropy And GA-BP Neural Network

As the aging problem of bridges becomes more and more serious in China,the damage identification and health monitoring of bridges have become an important direction in the future development of Bridges.Damage identification methods based on the theory of structural vibration characteristics and intelligent algorithms have made great progress,but there are still many difficulties in bridge damage identification.For example,it is difficult to obtain the measured complete information,the impact of environmental noise is great,and the quantitative analysis of structural damage degree is difficult.Therefore,based on the previous research and structural damage identification status quo,combined with information entropy,modal strain energy and intelligent algorithm theory advantages of structural damage location analysis and quantitative analysis,the relevant research is as follows:(1)The traditional modal strain energy theory is improved,the improved modal strain energy theory adopts the mode of vibration before and after the damage.Combined with the theory of improving modal strain energy and the theory of information entropy,it is derived from the function of improving modal strain energy entropy before and after injury,new damage identification indicators(Improved Modal Strain Energy Entropy Gradient,IMSEEG)are constructed to analyze the damage location of the structure.(2)Numerical simulation analysis of two different structural forms of bridges,simple support beam bridge and truss beam bridge is established,the effectiveness and accuracy of improving the modal strain energy change rate entropy(IMSEEG)index for damage location and the damage recognition ability under different degrees of noise are verified,and the damage degree of structure damage is analyzed by using traditional data fitting method.(3)Using the initial weight and threshold of BP neural network optimized by genetic algorithm,the GA-BP neural network is constructed,the IMSEEG damage identification index is used as the input of the network,and the stiffness reduction coefficient of each unit is output,so as to quantitatively analyze the structural damage.(4)To further verify the practical application ability of IMSEEG index and GABP neural network,a vibration test of simple steel beam bridge was carried out.The modal information before and after structural damage is obtained by using steel beam slotting to simulate structural damage,and the test data are processed and calculated step by step.Combined with theoretical analysis,numerical simulation and experimental verification,the full text proves the validity and accuracy of this method.The results show that the IMSEEG index derived from this paper can accurately locate the singlepoint damage and multi-point damage of the simple support beam and truss beam bridge.The IMSEEG index combined with GA-BP neural network can carry out accurate quantitative analysis of single-point damage and multi-point damage of structure,and the relative error can be maintained within 5%.IMSEEG can accurately identify the damage location in the environment of 20 d B and above when analyzing the damage of the bridge,and in the structure of the truss beam bridge,it is necessary to accurately locate the damage in the environment of 40 d B and above,indicating that the noise resistance performance of the IMSEEG index will be affected to some extent as the complexity of the structure increases.Finally,the accuracy of damage identification method is further verified based on a simply supported beam test,but the accuracy of damage identification is reduced by uncontrollable factors in the test.