DAMAGE IDENTIFICATION NEURAL NETWORK METHOD BASED ON STATISTICAL PROPERTY OF STRUCTURAL RESPONSES

A damage identification method using artificial neural network (ANN) based on a novel damage index, statistical property of structural dynamic responses, is proposed, and is evaluated through the numerical simulation and experiment verification. The feasibility of using the statistical property as damage index is validated theoretically with sensitivity analysis. The damage identification for a three-span continuous beam using the proposed method was numerically simulated, considering single damage case and multi-damage case. The results of numerical simulation show that the trained ANN can correctly identify the location and extent of damages in both single damage case and multi-damage case. A series of model tests of a fixed beam were performed to verify the validity and efficiency of the proposed method. From the results of test verification, it is shown that the trained ANN can correctly identify the location and extent of damage in single damage case and correctly detect the location of damage and mostly identify the extent of damage in multi-damage case. A conclusion is given that the novel method using the statistical property of structural response as damage index for damage identification is feasible and efficient.