The Complexity Of The Neural Network-based Structural Damage Identification

In the use of structure, it can not avoid that there are a lot of kinds of damage.The main two reasons are fatigue and air hole which produced in the welding process. These two reasons lead to some micro crack in the structure, under the alternate load,The concentrative stress will be produced around the micro cracks, and that lead to the cracks larger and larger. When the length of the crack comes to critical point, the component will lose stabilization and rupture suddenly, so that the structure invalidates.Damaged structure often leads to grave accident and results in the loss of life and property. So, the forecast of damage is very important. In recent years, the artificial neural network (ANN) as a new method has been developed rapidly and used widely. Because its high effective ability of nonlinear mapping, powerful abi I ity of solving contrary problem, abi I ity of compute in real time, it is an excellent unparameterized diagnostic method, thus the ANN is used in the field of damage identification. In this paper, a kind of new test parameter (flexibility) has been used as input of ANN, and substuctural method has beenused to analyse the complex structure.In study, a complex truss structure modeling is built and analysed by FEanalysis software (Ideas) in order to get the modal data. Then the flexibility is computed. The result is the input of the ANN. The samples of training andtesting are obtained by reducing stiffness to simulate the degrees of damage.The damage identification method of complex structure based on ANN is analysedand tested by substructural method.The result of experiment proves, as far as structure damage is concerned,the flexibility is more sensitive than the frequency and the mode. Theflexibility is the function of frequency and mode, it contains more information.So it is reasonable using the flexibility to test structure damage.