Study On Damage Identification For Single Pole Platform Structures Based On Frequency And Neural Network

Many studies of structure damage identification have done at home and abroad, much production and applied methods are achieved. Applicability and precision of the methods of damage identification to dissimilar structures are evidently discrepant in different working condition. Consequently, methods and instruments to structure damage identification will be selected according to the shape of the structure and its working condition.Offshore single pole platform is suffered from natural loads ,such as wave wind and sea ice and so on which due to bad condition. Single pole platform is important to the whole structure since it is the main bearing member of the structure. It is crucial that a suitable method of damage identification to single pole platform should be applied.In actual projects, especially to complicated configurations and huge structures, finite test points are definite. So the data extracted are limited. If other information is gained through extending the points, much error will be caused. In engineering testing, structure multiple frequencies are achieved by one or two testing points. Furthermore the data error of the modal vector often exceeds 20% in the test of the great civil engineering, but the error of the frequency doesn't exceed 1.5%. Consequently frequency method is suited to apply to the engineering.Based on this theory, this dissertation attempts to use the lower frequency and formulates the "frequency change square ratio" is formulated as structural damage sensitivity feature. Then it is the development of the method based on frequency. Firstly, this dissertation calculates the theoretical formula of the method of "frequency change square ratio". Secondly the FEM model of single pole platform is built. And then using decrease the Yang's modulus to simulate the damage, the damage can be located. The result is proved feasible. Thirdly, organic glass model of offshore single pole platform is built and the damage is simulated using rifting on the pole. And the "frequency change square ratio" is valid. So the method can be used in reality. Finaly, against the limitation of "frequency change square ratio" to the single pole platform, BP nerve network method is applied to damage identification of the structure based on FEM data. And the result is proved that this method is feasible.