Research On Anti-noise Performance Of Intelligent Calculation-based Damage Identification Method For Bridge

Under the influence of vehicle load, natural environment and material degradation, damage of existing concrete simply supported beam bridge with middle and small span continuously appears. Special shaped bridge serving as the main structure of urban overpass is unavoidable to be damaged due to the complicated mechanical characteristic. It equips with important practical significance to investigate the damage identification for concrete simply supported beam bridge and special shaped bridge. In addition, in the process of measuring the dynamic parameters of concrete bridge, the effect of noise is unavoidable. It leads to the missing of effective information involved in dynamic parameters, and makes the false damage identification result. The research on the anti-noise of damage identification method is necessary to improve the accuracy of damage identification approaches.In this paper, BP neural network(BPNN) and support vector machine(SVM) techniques are adopted to achieve the damage identification for concrete simply supported beam bridge and special shaped bridge, and verify its effectiveness of damage. In order to study the anti-noise performance of BPNN and SVM, the Gaussian white noise is added into the eigenvector for damage identification. Accuracy of BPNN and SVM for damage identification under different level of noise is investigated, which achieves the optimal selection for the damage identification based on BPNN and SVM algorithms. Specific works are carried out as follows:(1) The background and significance of investigation on the damage identification methods based on dynamic parameters and computational intelligence techniques are stated. And the advantage and disadvantage of existing damage identification approaches are comparatively analyzed. Analysis of anti-noise performance of damage identification methods is carried out, and helps to clear the important significance of studying the anti-noise performance. On the basis of summarizing the researches at home and abroad, the research ideas are proposed.(2) The principles of damage identification based on modal frequency and mode shape are systematically introduced. Moreover, the theoretical basis of BPNN and SVM are stated. It lays a foundation for the construction of damage identification system based on the computational intelligence techniques.(3) Damage identification for concrete simply supported beam bridge is carried out. BPNN and SVM are trained to identify the damage of simply supported beam bridge using the ratio of mode shape as the index of damage identification. Testing samples mixing with Gaussian white noise are selected to verify the anti-noise performance of BPNN and SVM algorithms.(4) Considering the special geometrical shape of special shaped bridge, the parameter of modal frequency is used to construct the damage identification index. Damage location and extent of special shaped bridge are identified by BPNN and SVM. The anti-noise performance of BPNN and SVM for special shaped bridge is checked by adding Gaussian white noise with different levels into the index vector for damage identification. Results of damage identification for testing sample are obtained and verify their anti-noise performances.