Reliability Design Of Complex Structure/Mechanism Based On Advanced Surrogate Models

Large-scale equipments such as aviation industry and mechanical engineering,and so on,are usually assembled by multiple complex structures/mechanisms according to a certain rule.These structures/mechanisms are often affected by multi-physical field,such as fluid,thermal,structure,etc.,in the working process,and these involved workloads have dynamic time-varying characteristics and uncertainties.It is significante to investagte the effects of the dynamic time-varying and uncertain parameters on output response,and to explore a reasonable and effective analytical models of reliability design,for improving the reliability of complex structures/mechanisms and ensuring the safe operation of large equipments.However,traditional surrogate models cannot effectively resolve the problem of high-nonlinearity,and cannot efficiently address the reliability design for dynamic time-varying characteristics and multi-disciplinary and multi-objective correlation.Around the engineering background of integrated reliability design of complex structures/mechanisms,this paper therefore focuses on the research of advanced surrogate models of dynamic probabilistic analysis(reliability and sensitivity analyses)and optimization design of multi-disciplinary and multi-objective.The detailed efforts are listed as follows:(1)Conventional response surface method(RSM)cannot effectively deal with the high-nonlinear problem,and cannot reasonably solve the reliability analysis with dynamic time-varying characteristics and multi-source uncertainties.To address the abovementioned issues,we develop weighted regression-based extremum RSM(ERSM)(WR-ERSM,short for)for reliability design of complex structures/mechanisms,which fuses the ERSM,weighted least square and fuzzy entroy principle.The effectiveness of the proposed method is verified using the reliability analysis of an aeroengine high-pressure turbine blisk.On this basis,an improved WR-ERSM(IWR-ERSM)is explored to further improve the accuracy of dynamic reliability design.An aeroengine low-pressure compressor blisk is treated as the study object to demonstrate the applicability of the developed method in reliability and sensitivity analyses of complex structures/mechanisms.(2)Trational Kriging model cannot accurately find hyperparameters for high-nonlinear,and cannot effectively derive the dynamic reliability design of complex structures/mechanisms.Thus,we develop an improved Kriging with ERSM(IK-ERSM)intergrating the ERSM,Kriging model and genetic algorithm,and select an aeroengine low-pressure compressor blisk as an example to validate the feasibility of the proposed method in dynamic reliability and sensitivity analysis.Furthermore,the application of extremum thought in the generalized regression neural network(GRNN)is discussed to develop an enhanced network learning method(ENLM),which is employed to complete the dynamic reliability analysis of complex structures/mechanisms.The practicability of this method is then illustrated combined with the dynamic probalistic analysis of low cycle fatigue life for an aeroengine high-pressure turbine blisk.(3)To overcome the problems that the modeling precision of RSM based on the least square method is unable to satisfy engineering requirements,and the solution procedures for the dynamic reliability design of complex structures/mechanisms are fussy,we probe a moving extremum surrogate modeling strategy(MESMS)based on the moving least square method and ERSM,which is utilized to investigate the dynamic reliability analysis of complex structures/mechanisms.On this basis,to avoid the premature phenomenon in the process of genetic algorithm optimization and improve the dynamic reliability prediction ability of IK-ERSM,a modified kriging-based moving extremum framework(MKMEF)is developed from the moving least square method,multi-population genetic algorithm,Kriging model and ERSM.Besides,an aeroengine high-pressure compressor blisk is taken as the study objective to verify the proposed method.(4)For resolving the dynamic reliability design of complex structures/mechanisms with assembly relation,we intergrate the decomposed-coordinated strategy into IK-ERSM and improved extremum Kriging surrogate model method-based least square,respectively,and propose two improved Kriging surrogate model methods based on decomposed-coordinated strategy including improved decomposed-coordinated Kriging modeling strategy(IDCKMS)and decomposed-coordinated framework with enhanced extremum Kriging(E2K-DCF),which are used to coordinate the relationship between the output responses of multiple components,and to study the dynamic probabilistic analysis of the singal failure mode of complex structures/mechanisms.Additionally,the effectiveness of these two methods is verified using the dynamic probabilistic analysis of radial deformation of an aeroengine high-pressure turbine blisk assembled by disk and blades.(5)To derive the dynamic comprehensive probabilistic ananlysis of complex structures/mechanisms and solve multip-failure modes correlation,we first propose a multiple ERSM(MERSM)based on ERSM.The application of decomposed-coordinated surrogate model method based on quadratic polynomial(QP-DCSMM)and decomposed-coordinated surrogate model method based on Kriging(K-DCSMM)in the dynamic comprehensive reliability analysis of multi-failure modes of complex structures/mechanisms is then explored.Finally,decomposed-coordinated surrogate model method based on mixture strategy(M-DCSMM)is proposed to further improve the accuracy and efficiency of dynamic comprehensive reliability analysis of complex structures/mechanisms.In addition,the feasibility of the developed methods is illustrated through the dynamic reliability and sensitivity analysis of an aeroengine high-pressure turbine blisk with multi-failure modes.(6)For the reliability-based optimization design of complex structures/mechanisms,IWR-ERSM is first used to derive single objective optimization design to deal with the problems of multi-disciplinary and a large number of cycle iterations.The effectiveness of the method in the optimization design is verified by the optimization design analysis of an aeroengine low-pressure compressor blisk stress.IDCKMS and MERSM based on multi-model are then applied to multi-objective coordinative optimization design and comprehensive optimization design,respectively,to resolve the multi-disciplinary and a large number of cycle nested iterations problems.Combined with the collaborative optimization design of high-pressure turbine blisk running casing and the comprehensive optimization design of high-pressure turbine blsk multi-failure modes including deformation,stress and strain,the feasibility of the proposed method for multi-objective optimization design is elabrated.