Ensemble Of Surrogate Models Based On Hybrid Measures And Its Applications In Finite Element Model Updating

In the harsh marine environment,corrosion,aging and accidental damage of offshore platform structures are inevitable.Once an accident occurs,it will cause tremendous economic,social and environmental impact.In order to ensure the reliability of complex structures and to avoid the occurrence of major malignant accidents,damage diagnosis and safety assessment of these structures must be carried out during their service period.To accurately evaluate the safety of large and complex structures,it is necessary to establish a reasonable finite element calculation model which can reflect the actual situation of the structure.However,the finite element model established according to the original design and completion data of the structure often deviates from the actual operation structure,including model simplification error,structural damage error and physical characteristics error,The commonly used method to solve this problem is the finite element model updating technology.In this thesis,the surrogate model technique is applied to the finite element model updating,The surrogate model is an approximate substitute model based on a small number of sample points.It can ensure the accuracy of calculation,while avoiding repeated invocation of the finite element model in the traditional parametric updating method,greatly reducing the calculation cost of the process of finite element model updating.In this thesis,a new unified ensemble of surrogates(Unified Ensemble of Surrogates,UES)is proposed,which further improves the fitting precision of the surrogate model.In view of the proposed method,this thesis mainly studies from the following aspects:(1)The principle and basic steps for the finite element model updating method based on the surrogate technique is introduced in detail.The common surrogate model,experimental design method for selecting sample points,evaluation criteria for fitting accuracy of surrogate model,selection principles of parameters to be updated and the optimization algorithm for parameter updating are also introduced.The advantages and disadvantages of the existing ensemble of surrogate technique is analyzed.In this thesis,a unified ensemble of surrogates technique(UES)is proposed.The method unifies the local error measures and the global error measures in the process of solving the weight coefficient.The control function is used to adjust the influence area of the two error measures when calculating the weight coefficients in the design space,the weight coefficients of each surrogate model can be evaluated more accurately,and the performance of ensemble surrogate model will be improved.(2)In order to evaluate the fitting performance of the proposed method comprehensively,a test function library consisting of 38 test functions with different characteristics is established in this thesis.The ensemble method is evaluated by three commonly used evaluation criteria for fitting accuracy of surrogate model.At the same time,the influence of the number of sample points on the fitting accuracy of the proposed ensemble method is further explored.(3)Two examples of finite element model updating are used to verify the application of UES in model updating.Firstly,the feasibility and reliability of UES in model updating are verified by finite element model updating of simply supported beam.Then eight parameters of offshore jacket platform are selected by sensitivity analysis for finite element model updating.And the difference between UES and other ensemble of surrogate technique in correcting accuracy is compared and analyzed.The results show that UES is superior to other ensemble of surrogate models in correcting accuracy.On the premise of guaranteeing the physical meaning of the modified parameters,the modified finite element model can reflect the mechanical properties of the structure more accurately.which lays a foundation for the safety assessment and structural health monitoring of the follow-up structure.In summary,the UES further improves the accuracy and reliability of the ensemble model technology.By fitting the test functions in the approximate function library,the advantages of the proposed method in fitting accuracy compared with other surrogate model technologies are verified.At the same time,it is a new attempt to apply the ensemble model technology to the finite element model updating of the offshore jacket platform.Two examples of finite element model updating prove the feasibility of the proposed method in the application of model updating.