Surrogate Model Technology And Its Application To Aircraft Reliability-Based Optimization

With the development of computer and aerospace technology,the analysis model of aircraft design is more and more sophisticated,and various uncertainties are also taken into consideration in the design.In order to obtain a more secure and reliable design scheme with high performance,designers gradually start to focus on Reliability-Based Optimization(RBO).However,facing the complex and time-consuming disciplinary analysis model of aircraft,reliability analysis process and optimization process nesting in reliability optimization face a huge challenge to solve.Therefore,in this paper,surrogate model technologies are used to reduce the amount of computation in these three areas.The main research contents and innovations are as follows:(1)A Cooperative Radial Basis Function(Co-RBF),a variable-fidelity model based on radial basis function,is proposed to reduce the computational complexity from the perspective of the analysis model.Co-RBF uses radial basis function and low-fidelity model as the basis functions,and constructs a new radial basis function,in which the model parameter is optimized by cross-validation.Finally,Co-RBF is compared with the other existing variable-fidelity methods through numerical and engineering examples.The results show that the proposed method is effective and efficient.(2)A global optimization algorithm based on Sequential Surrogate Based Optimization(SSBO)is proposed to reduce the computational complexity of the time-consuming model from the perspective of optimization algorithms.The gradient-based optimization algorithm is fast but the global searching ability is weak.The global searching ability of the heuristic algorithm is strong but the convergence speed is slow.In SSBO,surrogate models are used to approximate the objective and constraint functions,and traditional heuristic algorithm is adopted to search for possible areas containing global optimal points for incrementing points and constantly update the surrogate model.The hybrid criterions of adding points contain global and local searching ability,therefore the proposed method has strong global search ability and fast local convergence rate.Numerical and engineering examples show that the proposed method can converge to the global optimum quickly for the local multi-peak problem.Compared with the existing global optimization algorithms,local optimization algorithms and other similar optimization algorithms based on the surrogate model,the global convergence ability and convergence speed of SSBO have significant advantages.(3)A Sequential Surrogate Reliability Method(SSRM)is proposed to reduce the computational complexity from the perspective of reliability analysis.SSRM constructs the surrogate model of the limit state function sequentially.Through the conventional constrained optimization method,the surrogate model is searched for the failure boundary of the state function and the sample point with larger failure probability to improve the accuracy of the local important region.Then,the Monte Carlo simulation(MCS)based on the surrogate model is performed to obtain the failure probability,repeate the process of adding points and Monte Carlo simulation until the algorithm meets the termination conditions.Finally,results of numerical examples and engineering examples are compared with the existing reliability methods to verify the accuracy and efficiency of the proposed method.(4)A Sequential Surrogate Reliability-Based Optimization(SSRBO)method is proposed to reduce the computational complexity of the reliability analysis and optimization algorithm in the reliability optimization by using the surrogate model technology.SSRBO is divided into two stages.In the first stage,the sample points are added near the deterministic optimal point through the point-increasing method of deterministic optimization,and the surrogate models of the objective and the constraint function are respectively constructed.In the second stage,the surrogate model of the constraint function is implemented by Monte Carlo,and then the inverse cumulative probability density function is fitted through the simulated sample points to further obtain the constrained boundary shifts,thereafter the constraints shifted optimization problem is solved.The process is repeated until the shifts converge.The proposed SSRBO algorithm can significantly reduce the number of evaluations of the time-consuming model,and transform the calculation to the relatively fast surrogate model optimization and Monte Carlo calculation.Finally,numerical examples are used to verify the algorithm,which shows that the algorithm has the advantage of computational efficiency and accuracy in comparison with the existing reliability analysis methods.(5)The proposed method is applied to the optimization design of multidisciplinary reliability of suborbital flight vehicle.Firstly,the task of returning for the suborbital flight vehicle is analyzed,and the basic scheme and disciplinary model of the multidisciplinary design problem are established.Then the deterministic optimization based on the basic scheme is carried out.Then the reliability analysis is carried out at the optimal point.Finally,uncertainties are introduced into the larger design variables and reliability-based optimization is performed.In summary,this paper aims at solving the problem of large computational complexity in RBO of complex time-consuming models.Methods to reduce the computational burden from different perspectives are studied,including the model analysis,reliability analysis and global optimization.Finally,the proposed method is applied to the multi-disciplinary reliability optimization problem of suborbital flight vehicle.The feasibility and effectiveness of the proposed methods in engineering application are verified,which lays a foundation for the development of reliability optimization and the improvement of aircraft reliability design.