Research On Structural Reliability Optimization Method Based On Model Fusion

Structural reliability-based optimization design usually involves the solution of implicit function.Generally,it is necessary to build a proxy model to replace the real model for calculation.When modeling,it needs to call a large number of highprecision models to ensure the accuracy of the model,which often leads to a waste of computing resources.The model fusion technology can effectively fuse the data of high-precision and low-precision models,and improve the modeling efficiency while ensuring the model accuracy.Therefore,the model fusion technology is of great significance in the structural reliability-based optimization design.In the past,it is difficult to predict the number of sample points in advance to ensure the accuracy of the model.Therefore,this paper proposes a model fusion method based on sequence sampling and Gaussian process,and develops a reliability-based optimization design method based on local updating fusion model.The main tasks are as follows(1)A model fusion method based on sequence sampling and Gaussian process is proposed to realize the effective fusion of high and low precision model response data.The initial sample points are obtained by Latin hypercube sampling to construct the initial fusion model.On the basis of the initial fusion model,the accuracy of the new sample points is judged based on the sequence sampling criteria,and the fusion model is updated step by step.The example shows that the fusion model updated by the sequence sampling criterion has high prediction accuracy.(2)A reliability-based optimization design method based on local updating fusion model is proposed,which realizes the reliability-based optimization design of response evaluation by using fusion model in both objective function and constraint condition.In reliability-based optimization design,the double nested problem is decoupled into design variables optimization and reliability evaluation by decoupling method.The initial fusion model is established for the objective function and constraints respectively.During the decoupling process,the fusion model is updated by sequential sampling at the maximum possible failure point and the optimal solution.The example shows that the fusion model has better prediction accuracy when applied to reliability optimization,and the results of design optimization meet the requirements of reliability.(3)The reliability-based optimization design method of structural system based on local updating fusion model is developed to solve the reliability-based optimization design of structural system under multiple failure modes in practical engineering problems.The initial Gaussian process model is established for the objective function and the constraint conditions under multiple failure modes.The uncertainty problem is transformed into a deterministic problem by decoupling method,and the fusion model is updated synchronously in reliability optimization.The example shows that the structural system reliability-based optimization design based on local updating fusion model can ensure the prediction accuracy and reduce the call times of highprecision model.