Damage Diagnoses Of Steel Grid Structure

With the growing popularity of the application of steel grid structure at home and abroad,accidents of such structure are increasing. As this structure has a large span, there are manypeople when it is constructing and serving. As a result, the accident will threaten human livesand cause huge economic losses and property loss. Structural damage identification is theresearch to avoid the occurrence of such problems. According to the knowledge of structuraldynamics, any structure can be regarded as a mechanical system which is constitutive ofstiffness, mass and damping. So any damage of structure will definitely lead to the change ofstructural stiffness, damping, consequently, result in the changes of frequency responsefunction and modal parameters (frequency and mode, etc.). Therefore, the key to damagediagnoses is how to relate the changes of modal parameters with structural damage.In order to study the various damage diagnosis methods, the finite element method hasbeen used to found orthogonal quadrangular pyramid steel grid numerical model. Throughcalculation the state of intact and damaged structure, the change of dynamic parameters hasbeen gotten. These parameters mainly refer to the structure of the natural frequency andvibration mode. And modal strain energy has been induced to conduct preliminary damagedetection.BP neural network model has been founded to diagnose structural damage. Based onprincipal component analyses and some other mehods, modal information has beenappropriately used as input of artificial neural network (ANN), and different training methodshave been compared. At last, a well trained network has been used to identify structuraldamage.Meanwhile, statistical learning theory has been introduced to conduct damage diagnoses.Statistical learning theory has founded the theoretical framework and common methods aboutlearning machines with limited sample. And support vector machine (SVM) is a learningmachine which is based on this theory. It has both rigorous theoretical foundations, and canbe a very good solution to the small sample, nonlinear, dimension disaster and local minimaand other practical problems. In this paper, this method has been induced to damageidentification. And least squares support vector machines (LS-SVM) has been used toaccurately detection damage location and evaluate the grade of damage of steel grid structure.According to the result of numerical examples, the conclusion can be drawn that SVM is agood method to damage identification.