New Methods Of Uncertainty Analysis And Design Optimization And Their Applications On Thermal Protection System

Today,most of mechanical and structural products are still designed by simplified rules or strategies,such as safety factor.However,these traditional design methods are based on experiences and engineering judgment so that they can not consider the random features of the inputs and result in conservative designs.As the requirements in modern engineering products on key and complex designs increase,methods,which can efficiently and accurately evaluate the uncertainties in computational models,loads,geometry features,material properties,manufacturing and operational environment,are highly demanded.It is well known that uncertainty and optimality are major considerations in structural design.On one hand,the thesis handles with the combined size and shape optimization of tuss structures by Subset Simulation Optimization(SSO)for both continuous and discrete design variables.To demonstrate its search capacity,four numerical examples are considered,including a 15-bar 2D truss,an 18-bar 2D truss,a 39-bar 3D truss and a truss-type landing gear of an unmanned aerial vehicle.On the other hand,this thesis uses a posterior approximation approach with a recently developed Generalized Subset Simulation(GSS)method to approximate the probabilistic constraints in Reliabiltiy-Based Design Optimization(RBDO)problems.Based on this,the double loop RBDO is transformed into a single one and can be solved by any kinds of optimization methods.Besides,the performance of this approach will be tested by three typical examples.Based on the above methods,this thesis carries out an uncertainty analysis on the non-ablating Thermal Protection System(TPS)during the conceptual design phase.This is implemented by means of GSS and a space-partition Monte Carlo sensitivity approach.The uncertainty analysis mainly focuses on uncertainty characterization,definition of failure modes,failure probability evaluation,global sensitivity analysis(based on the first-order sensitivity index),and model reduction(based on global sensitivity analysis).To be more specific,the uncertainty anlaysis procedure can be listed as follows.Firstly,TPS selection and sizing using sequential quadratic programming design optimization are operated firstly to provide the nominal values including allowable temperature limits and thicknesses of TPS materials as the distribution parameters of the uncertainty inputs.Secondly,multi-inputs and multi-outputs support vector machines are utilized to approximate the thermal responses when failure modes are constructing,for the sake of dramatically reducing the amount of calculations.Thirdly,two application examples consisting of a lifting body vehicle model and a spacecraft model for conceptual design are used to demonstrate the performance of the proposed uncertainty analysis approach.Fourthly,the proposed RBDO method is tested on nodes where the associated failure probabilities are relatively large in both the lifting body model and the spacecraft model.Finally,SSO is also considered for TPS deterministic optimization that combines TPS selection and TPS sizing together.