Bridge Damage Detection Based On Modal Analysis Theory And Improved BP Neural Network

Based on the study mode theory and neural networks, through the ANSYS and MATLAB, the paper proposed the new identification method which use the new flexibility curvature index D_FC and the neural network to identify damage of the simply beam bridge,there are four part of this paper.First, the damage identification method based on vibration modal analysis theory is discussed and analyzed. During the process, natural principle and detected procedure of this analysis theory are introduced. natural principle, predominance and shortcomings of every capabilities and applicable scope about modal analysis methods are systematically analyzed.Second, the damage identification method based on BP neural networks is discussed and analyzed. This part introduced the basic theories and mathematic deduction of artificial neural networks. Then this part introduced BP neural networks, BP classic algorithm and the Lavender-Marquardt algorithm.Third, through the numerical examples of a simple beam bridge, established damage identification methods which based on D_FC and neural network. The simply beam identifies only one unit who has damage and identifies two units who has damage. Then identifies the simply beam's damage units when the number of the damage units is unknown. The identification of the damage location is completely correct, and the relative error of the damage degree also can be accepted.Lastly, when using the method which uses the combination of D_FC indicator and BP neural networks to identify the damage of the bridge model of Zhuanshanzi, the paper proposed a new method which has two steps, and the first step analyses the vertical location and the second step analyses the lateral position. So the calculation is greatly reduced, and make the method which uses the combination of D_FC indicator and BP neural networks to identify the damage becomes possible to be applied in the actual bridges. The two-step method which bases on D_FC indicator and BP network achieved good results in the bridge model of achieved. Identified the vertical location of the damage object, and then identified the horizontal position of the damage object. The degree error of injury within the acceptable range also.