Research On Modal Parameters Of Bridge Structure: Identification And Application

Long-span bridges are one important part of state basic facilities. In the period of bridge's service, different degree levels of damages will occur after suffering to various natural disasters and human artificial destroying, which will not only increase gradually the aging of the bridges and debase the load-carrying capacity capability of bridge structures, but also affect the safe operating of bridges. Therefore a predicting and warning and timely maintaining system based on identifying the modal parameters and inspecting damages of the bridge structures is developed which can help to eliminate hidden troubles and avoid the occurrences of catastrophic accidents.Based on the study on Ligong first bridge of Wuhan University of Technology, this thesis starts from the finite element theoretic model of the bridge, then the whole bridge health monitoring process is studied roundly and systemically, including the dynamic characteristic testing, modal parameters identification, structural damages detecting. Some theories and methods are applied in this paper, including finite element simulation, system identification, modal parameter identification and artificial neural network. The main research works and innovations of this paper are as following:Firstly, the finite element theoretic model of the Ligong first bridge is constructed and the dynamic characteristics of the bridge are studied based on this model in order to master the changing rules of the bridge's eigenfrequecies and mode shape. The results are exact and credible, which provide theoretical basic for further research on bridge operating state.Secondly, the dynamic characteristics of Ligong first bridge are tested from ambient excitation. Seven questions needing attention which affect directly the accurancy of test are raised up. Then its important structure modal parameters including eigenfreqencies, damping ratio and mode shape, are identified by applying admittance circle fitting method. This method can overcome the shortcoming of the method of finite element which is not capable of computing the damping ratioThirdly, a method of reference based stochastic subspace identification (SSI/REF) is discussed. In this method, the output measurements from ambient excitation are gathered directly in a block Hankel matrix. Then SSI reduces data through QR factorization and applies SVD to eliminate noise, at last the structural modal parameters are identified. SSI can't affected by FFT, which is one of time domain method with fast computation rate and high identification accuracy. Comparing the result of the SSI/REF method with the admittance circle fitting method, the SSI/REF identified results are closed to theoretical results. Under the circumstance that the test results are exact and error disturbances are small, the damping ratio identification has high creditability.The forth, the finite element model modification method is studied. Firstly, the model is modified by using cable tensions of the bridge, which are tested based on ambient vibration testing method. When the cable tensions is evaluated by applying the natural frequencies obtained, the effects of sag, slope, bending stiffness and anchoring method on the cable tension are fully considered. It shows that the cable of Ligong first bridge can be assumed rigidity cable. Through comparing the measured value with designed one, the problem of the cable of this bridge was found. Secondly, the finite element model modification method using neural network is studied. This neural network is trained by computing sample of finite element model, then the modal parameters identified by SSI is used to test the neural network. After modifying, the model is closed to the real one, which demonstrates that the method is effective.The fifth is to study the bridge structure damage identification based on neural network. Firstly, the natural frequencies' changing rules of the girder, tower and cable of the Ligong first bridge suffered to different levels of damages with the changing of elastic module are simulated by finite element. It shows that the natural frequencies of main girder will change more obviously than the main tower and the cable following the reducing of elastic module, and their sensitivity of damage at high rank frequencies is higher than that at low rank frequencies. Secondly, the different places and different levels of the bridge damages are located and identified precisely by neural network. This method has higher precision and little error.The sixth, as a result of the above studies, computational software of reference based stochastic subspace identification is developed, which is tested by the measuring data of Ligong first bridge. It turns out that the program can not only reduce the large datum effectively, but also make the identifying results close to the measuring results, which demonstrates that the method is one advanced identification method.