Research On Reliability Robust Optimization Design Method Based On Gauss Process Model

Uncertainty generally exists in practical engineering problems.With the use of design optimization method under uncertainty,there are increasing demands for engineering designs to have higher reliability,higher robustness,lower risk as well as lower cost,which is of great significance to engineering design.Nowadays,the design of uncertain conditions has been widely used in various fields,such as aerospace engineering,mechanical engineering and civil engineering.It usually includes reliability optimization design and robust optimization design.Simulation models are created to replace usually extremely expensive physical experiments,however,due to the lack of knowledge and other factors,there is a lot of uncertainty in the simulation model.The uncertainty of the model is the main source of uncertainty.Therefore,the uncertainty of the model should be considered when uncertain optimization is carried out.However,due to the fact that cognitive uncertainty is related to human subjectivity and it is difficult to use unified standard description.Most of the traditional optimization design only takes into account the stochastic uncertainty.The uncertainty of the model is seldom taken into account in the optimization design.In this paper,the Gauss process model is used to measure the uncertainty of the model.On this basis,the reliability optimization model and the robust optimization model are established at the same time of the stochastic uncertainty and the cognitive uncertainty.The sequence optimization method and the SORA are proposed to decouple the nesting problems in the model and improve the computational efficiency.Through the Monte Carlo method,the model uncertainty is introduced into the inner layer reliability calculation,and the reliability analysis of the inner layer and the design optimization of the outer layer are continuously circulate until the final convergence,thus the reliability optimization and the robust optimization design consideration of the model uncertainty are obtained.The specific research contents of this paper can be divided into three aspects:(1)Through the Gauss stochastic process,the response information of different precision models is effectively and reasonably fused to solve the problem that the cognitive uncertainty of the model is difficult to describe accurately,and the accurate measurement of the model uncertainty is realized.Aiming at the problem that the high precision data is difficult to obtain,and the low precision model leads to the low reliability.We use the Gauss Process model to integrate the sample data of different precision effectively and reasonably.The GP model can be guided to ensure the precision of the model with a few high precision data,which effectively overcomes the dependence of the traditional modeling method on high precision data and avoids the low efficiency.It also avoids the low accuracy of data due to the lack of data.The Gauss process model gives not only the corresponding mean of output,but also the mean variance of the output response,which shows the confidence of the output,so as to realize the accurate measurement of the uncertainty of the model.(2)On the basis of establishing the Gauss process model,the uncertainty of the model will be considered in the reliability optimization design,a reliability optimization model based on the existence of stochastic uncertainty and cognitive uncertainty is established and the problem of its solution will be studied.Aiming at the double nesting problem of the reliability optimization model,the SORA method is introduced to decouple so as to improve the computational efficiency.In view of the cognitive uncertainty in the reliability analysis stage,the Monte Carlo method is used to introduce the mean and variance of the response in the Gauss process model to the calculation of reliability,so that the reliability is reached.In view of the cognitive uncertainty in the reliability analysis stage,the Monte Carlo method is used to introduce the mean and variance of the response in the Gauss process model to the calculation of reliability,so that the reliability meets the requirements.The sequence is iterated through deterministic design optimization and reliability analysis until final convergence.The result of reliability optimization is obtained when both random uncertainty and cognitive uncertainty exist simultaneously.(3)On the basis of establishing the Gauss process model,reliability optimization and robustness optimization are combined in the process of optimization design and the problem of its solution will be studied.By combining the two basic methods of uncertain optimization design:the reliability optimization design and the robust optimization design,the uncertain optimization design not only takes into account the reliability of the structure,but also prevents the instability of the structural performance resulting from the change of the parameters.Aiming at nesting problem,we use sequence robust optimization method based on mobile constraint function boundary and SORA method to decouple the model to improve the computational efficiency.Through the Monte Carlo method,the model uncertainty is introduced to the reliability analysis,and the result of reliability robust optimization is obtained when both random uncertainty and cognitive uncertainty exist simultaneously.