The Research On Structure Damage Identification Method Based On Multi-modal Parameters

Structure dynamic damage identification is a research field with a good perspective. In this paper, the method of using the structure multi-modal parameters as the network inputs in the damage diagnosis has a definite physical meaning, especially that measuring few modal steps to make the engineering application easily, both the theory and method are suitable to the average engineering structure.Consider the fact that natural frequency can reflect the structure entity performance but cannot be sensitive to the structure local damage identification. In many cases, structure damage in deferent location can cause almost the same decrease in low frequency, so this method cannot ascertain the definite damage location. Through contrast, mode data are more sensitive to structure local damage than frequency, but it has the question that the measured accuracy is lower than natural frequency which make results having big deviation. So the selection of input parameters affects the identification performance directly, In this paper, using the integrate parameters of frequency and mode as the network input parameters to implement damage identification. Through compared with the method using single frequency or mode, this method may make up the difficulties caused by vibration reverse questions.In this paper, ANSYS data model and corresponding physical model are built in simple concrete beam based on the theory and method of structure diagnosis and made a experiment, In order to examine the diagnosis ability of diagnosis theory, many kinds of diagnosis designs are put forward. At the same time, using ANSYS to build data model to make analysis computing, using shaft net and concrete net to build ANSYS model to have the corresponding integrate before and after the damage, and using suitable implored BP neural network to ascertain the damage location and extent. The results is that the neural network applied in structure is feasible and have good application perspective. The neural networks are more and more important in engineering applications, especially in forecasting building healthy state, In addition, this method has the features of simple, rapid, no damage and has good economical and social advantages.