Study On Methods Of Key Computation And Software Platform Of Structural Health Monitoring For Cable-stayed Bridge

For large-span bridge structures such as new and existing cable-stayed bridges,effective means must be adopted to monitor and evaluate their working condition and damage status.As a result,the health monitoring technology came into being and turned into a research hotspot.Due to the complexity of cable-stayed bridges and other large-span bridge structures,some limitations and bottlenecks in application of the existing technology,existing structural health monitoring technology was not mature,and many difficulties had been encountered in practical applications,many key computing problems need to be solved urgently.With the cable-stayed bridge as the object,the methods of key computation for health monitoring of cable-stayed bridges were systematically studied.The research completed and results achieved mainly include the following six aspects:(1)The cable-stayed bridge health monitoring software platform was developed with the VC ++.NET development platform and two kinds of engineering software—ANSYS and MATLAB;nine computation modules were integrated,including the structure static calculation module,structural modal calculation module,time course calculation module,test design module,fingerprint calculation and sensitivity analysis module,test data preprocessing and modal parameter identification module,response surface calculation module,optimization calculation module,and pattern recognition module.On this platform,the research on the methods of key computation for cable-stayed bridge health monitoring was conducted.(2)In light of some drawbacks of the conventional methods at the time of establishment of a baseline finite element model for cable-stayed bridge health monitoring,this paper proposed a new method in combination with the characteristics of cable-stayed bridges;demonstrated the differences and relations between the new method and the existing similar method and theory,deduced the basic theory of the new method,obtained the implementation steps of the new method,gave the implementation code of the key links;introduced the outline of the cable-stayed bridge project on which the paper was based;exhibited the realization process of the new method by depending on the cable-stayed bridges,comparatively verified the results before and after optimization and results of optimization and completed bridge tests,and demonstrated that the method proposed in this paper was necessary and effective for the establishment of the baseline finite element model for health monitoring of such specific bridge type such as a cable-stayed bridge.(3)Starting with the damage identification requirements of optimal sensor location,the author introduced the test design and parameter correlation analysis theory,proposed and obtained the node freedom damage information index containing all bridge section damage information and the method for optimal sensor location;defined the new index,elaborated the basic connotation of the new method,performed the theoretical derivation,and derived the realization process of the new method;through depending on the cable-stayed bridge project,presented the realization process of the proposed method,with result analysis proving the superiority of the proposed method.(4)Two recognition algorithms of NExT-SSI and FRF-PolyMax were introduced and implemented;based on the role of true and false modal screening in the modal parameter identification of ambient excitation,with regard to some problems with existing true and false modal screening methods,the paper put forward a new method of modal true and false screening by combination of stability diagram approach and power spectrum summation peak method,demonstrated the necessity and feasibility of the new method,and deduced the realization methods and steps for the new method,with a new method realized by programming;performed exemplary presentation of the modal parameter recognition algorithm and the proposed new method of true and false modal screening by relying on the cable-stayed bridge project.The results showed that the identification algorithm of the FRF-PolyMax method had higher precision and proved the necessity and effectiveness of the proposed new method of true and false modal screening.(5)In the application of response surface,the author proposed a new method of direct response surface optimization for damage identification;expounded the difference between the new method and the existing response surface method for bridge damage identification,and defined the concept of the new method,compacting the basic thinking,stating the three key problems in the application of this method,revealing the advantages of the new method,obtaining the flow chart for realization of the new method,discussing in detail the implementation steps of the new method and implementing the programming realization;introducing and realizing the four response surface models including the polynomial model,Chebyshev model(orthogonal polynomial model),radial basis function(RBF)model,and Kriging model and two multi-objective optimization algorithms including NSGA-II and particle swarm optimization algorithm;the detailed exemplary presentations and result analysis were given with regard to the proposed method and algorithm,and the effectiveness was verified by the measured data.(6)It was pointed out that the error problems with cable-stayed bridge monitoring data could not be solved simply by means of the accuracy of pattern recognition algorithm,but the method of establishing a fingerprint database should be studied deeply,and a new method of establishing the fingerprint database was put forward and realized.The paper compacted the basic thinking of the new method,pointed out that the core problems of the method were in two parts: the full-factorial design and the random error,explained deeply the two parts,named such sample database as "full-factorial design sample database containing " colored " errors",revealed the innovative features of the proposed method,and clarified the precautions for applying the method;elaborated the concrete implementation steps of the new method,and gave a realization flow chart of the new method;introduced and implemented the 4 model classification algorithms of high recognition accuracy including the multi-layered perceptron bagging algorithm,random forest algorithm,k-nearest neighbor algorithm and LMT decision tree algorithm;the detailed exemplary presentations and result analysis were given with regard to the proposed method and algorithm,and the effectiveness was verified by the measured data.