Research On Multi-alternatives For Finite Element Models Of Bridge Structures Based On Group Intelligence Algorithm

The traditional finite element model updation technology uses an efficient intelligent algorithm to find a set of global optimal solutions for the theoretical response and measured response error objective functions,so that the objective function value is minimized.However,due to the limited number of sensors,insufficient measurement data,instrument identification errors,and uncertainty in the construction of the objective function,there may be multiple sets of solution sets with the same or similar objective function values,but their solution spaces are quite different.Therefore,it is unreasonable to rely on a set of global optimal solutions as the finite element model that best represents the actual parameter values of the structure.There is an urgent need for multiple sets of global optimal solutions or local optimal solutions that can effectively reduce the error function based on comprehensive model analysis technology,decision makers combined engineering experience and reanalysis techniques to select one or more groups from multiple sets of solutions to reduce the possibility of misjudgment of structural parameters.Based on the above objectives,this paper proposes an improved adaptive niche fish school algorithm.This algorithm combines niche technology and introduces a reverse learning strategy.The algorithm is executed in conjunction with the simulated annealing algorithm for detailed optimization,which makes the improved algorithm can quickly and accurately find multiple sets of global optimal solutions of functions,and it can also effectively find local optimal solutions of multi-peak functions.Aiming at the difficulty of determining the niche radius by niche technology,this paper proposes an adaptive niche radius mechanism.The effectiveness of the improved algorithm is verified by numerical simulation analysis and used to study the multialternatives problem of finite element model updation.First,in the process of modifying the ASCE-Benchmark frame finite element model,through artificially presetting damage,and then modifying the damaged structure finite element model,a group of global optimal solutions and 12 groups of local optimal solutions were obtained.Comparing the preset damage value with the modified solutions,it is found that the corresponding parameter values of the global optimal solution are different from the preset values,but the sub-optimal solution 1 is in good agreement with the preset value,and the global optimal solution is demonstrated the view that may not be the most representative of the actual parameter values of the structure illustrates the significance of the research on the modified multisolution problem of the finite element model.It also proves that the improved algorithm can be applied to the research of the modified multi-solution problem of the complex finite element model.Secondly,the proposed adaptive niche fish school algorithm combined with polynomial response surface technology is applied to a more complex scaled cablestayed bridge model updation.Based on the measured data,stochastic subspace identification is used to identify the modal frequency of the bridge,and the parameters to be modified are selected in combination with engineering experience and the F-value test method in mathematical statistics.A quadratic polynomial response surface model with cross-terms with modified parameters as input and modal frequency as output was constructed,and the accuracy of the proxy model was tested.Using the proposed improved algorithm,eight sets of solution sets were found which effectively reduced the theoretical calculated and measured values of the scaled cable-stayed bridge,and the solution space was very different.Four representative sets of solutions were selected.Decision makers can weigh the pros and cons based on engineering experience,choose from multiple sets of solutions,and reduce the possibility of misjudgement of actual structural parameters.Finally,in order to further study the multi-alternatives selection method in the finite-element model updation multi-alternatives problem,the proposed adaptive niche fish school algorithm combined with the Kriging proxy model was applied to a more complex cable-stayed pedestrian bridge model updation.Based on the measured data,stochastic subspace identification is used to identify the modal frequency and mode information of the bridge,and the parameters to be modified are selected based on the analysis of the actual bridge conditions and parameter sensitivity.The Kriging response surface model of the output measured MAC value is used to verify the accuracy of the response surface model.The objective function was constructed based on the structure’s measured frequency,mode data,and theoretical value errors.Multiple sets of solution sets were found that reduced the value of the error objective function and had significantly different solution spaces.Four representative sets of solutions were selected.Based on the author’s shallow engineering experience and re-analysis technology,a set of solutions that can best represent the parameters of the real bridge are selected.The other sets of solutions can be used for the reliability analysis and risk assessment of the bridge.