To address the limitations of existing coupling methods in aero-engine system simulation,which fail to adaptively adjust iterative parameters and coupling relationships,which can result in low efficiency and in⁃stabil...To address the limitations of existing coupling methods in aero-engine system simulation,which fail to adaptively adjust iterative parameters and coupling relationships,which can result in low efficiency and in⁃stability,this study introduces a‘Dynamic Event-Driven Co-Simulation’algorithm integrated with decision tree algorithms.This algorithm separates the overall coupling relationships and the main solver from the primary mod⁃el,utilizing a dynamic event monitoring module to adaptively adjust simulation strategies,including iteration pa⁃rameters,coupling relationships,and convergence criteria.This facilitates efficient adaptive simulations of dy⁃namic events while balancing solution accuracy and computational efficiency.The research focuses on a twinshaft turbofan engine,establishing six system-level models that encompass overall performance and various sub⁃systems based on three coupling methods,along with a multidisciplinary multi-fidelity simulation framework in⁃corporating a 3D CFD nozzle model.The study tests both model exchange and coupled simulation methods under a 14 s transient acceleration and deceleration scenario.In a 100%throttle condition,a high-fidelity nozzle model is used to analyze the sensitivity of different convergence criteria on computational efficiency and accuracy.Re⁃sults indicate that the accuracy and efficiency achieved with this method are comparable to those of PROOSIS soft⁃ware(18 s and 35 s,respectively),while being 71%more efficient than Simulink software(62 s and 120 s,re⁃spectively).Furthermore,appropriately relaxing the convergence criteria for the 0D model(from 10-6 to 10-4)while enhancing those for the 3D model(from 3000 steps to 6000 steps)can effectively balance computational accuracy and efficiency.展开更多
The resource-intensive,high-fidelity infrared signature simulations and Radar CrossSection(RCS)calculations limit the integrated optimization of Unmanned Combat Aerial Vehicles(UCAVs)in response to escalating threats ...The resource-intensive,high-fidelity infrared signature simulations and Radar CrossSection(RCS)calculations limit the integrated optimization of Unmanned Combat Aerial Vehicles(UCAVs)in response to escalating threats from joint detection systems.To this end,we present a sample-efficient framework to advance the optimization efficiency of UCAV's exhaust system,focusing on both the stealth characteristics evaluation and the optimization process.A novel multi-fidelity stealth assessment method,powered by multi-fidelity neural network and local perceptive fields,has been developed to fuse different fidelity information from infrared radiation signature and RCS values,respectively.Results demonstrate that the method can achieve relatively high accuracy based on a small set of high-fidelity data.Furthermore,this data fusion method is integrated into a multi-objective Bayesian optimization framework.Employing a Gaussian process regression model and the EHVI acquisition function,the framework effectively explores the stealth objective space,achieving a 15.21%hypervolume indicator increase with fewer optimization iterations compared to NSGA-Ⅱ.Results show that the optimized nozzle significantly reduces both the infrared signature and RCS compared to the baseline configuration.The proposed framework offers a practical and efficient approach for optimizing the integrated stealth performance of UCAVs.展开更多
In the realm of subsurface flow simulations,deep-learning-based surrogate models have emerged as a promising alternative to traditional simulation methods,especially in addressing complex optimization problems.However...In the realm of subsurface flow simulations,deep-learning-based surrogate models have emerged as a promising alternative to traditional simulation methods,especially in addressing complex optimization problems.However,a significant challenge lies in the necessity of numerous high-fidelity training simulations to construct these deep-learning models,which limits their application to field-scale problems.To overcome this limitation,we introduce a training procedure that leverages transfer learning with multi-fidelity training data to construct surrogate models efficiently.The procedure begins with the pre-training of the surrogate model using a relatively larger amount of data that can be efficiently generated from upscaled coarse-scale models.Subsequently,the model parameters are finetuned with a much smaller set of high-fidelity simulation data.For the cases considered in this study,this method leads to about a 75%reduction in total computational cost,in comparison with the traditional training approach,without any sacrifice of prediction accuracy.In addition,a dedicated well-control embedding model is introduced to the traditional U-Net architecture to improve the surrogate model's prediction accuracy,which is shown to be particularly effective when dealing with large-scale reservoir models under time-varying well control parameters.Comprehensive results and analyses are presented for the prediction of well rates,pressure and saturation states of a 3D synthetic reservoir system.Finally,the proposed procedure is applied to a field-scale production optimization problem.The trained surrogate model is shown to provide excellent generalization capabilities during the optimization process,in which the final optimized net-present-value is much higher than those from the training data ranges.展开更多
Constructing high approximation accuracy surrogate model with lower computational cost has great engineering significance.In this paper,using co-Kriging method,an efficient multi-fidelity surrogate model is constructe...Constructing high approximation accuracy surrogate model with lower computational cost has great engineering significance.In this paper,using co-Kriging method,an efficient multi-fidelity surrogate model is constructed based on two independent high and low fidelity samples.Co-Kriging method can use a greater quantity of low-fidelity information to enhance the accuracy of a surrogate of the high-fidelity model by modeling the correlation between high and low fidelity model,thus computational cost of building surrogate model can be greatly reduced.A wing-body problem is taken as an example to compare characteristics of co-Kriging multi-fidelity(CKMF)model with traditional Kriging based multi-fidelity(KMF)model.A sampling convergence of the CKMF model and the KMF model is conducted,and an appropriate sampling design is selected through the sampling convergence analysis.The results indicate that CKMF model has higher approximation accuracy with the same high-fidelity samples,and converges at less high-fidelity samples.A wing-body drag reduction optimization design using genetic algorithm is implemented.Satisfying design results are obtained,which validate the feasibility of CKMF model in engineering design.展开更多
Front Variable Area Bypass Injector(Front-VABI) is a component of the Adaptive Cycle Engine(ACE) with important variable-cycle features. The performance of Front-VABI has a direct impact on the performance and stabili...Front Variable Area Bypass Injector(Front-VABI) is a component of the Adaptive Cycle Engine(ACE) with important variable-cycle features. The performance of Front-VABI has a direct impact on the performance and stability of ACE, but the current ACE performance model uses approximate models for Front-VABI performance calculation. In this work, a multi-fidelity simulation based on a de-coupled method is developed which delivers a more accurate calculation of the Front-VABI performance based on Computational Fluid Dynamics(CFD) simulation. This simulation method proposes a form of Front-VABI characteristic and its matching calculation method between it and the ACE performance model, constructs a coupling method between the(2-D) Front-VABI model and the(0-D) ACE performance model. The result shows, when ACE works in triple bypass mode, the approximate model cannot account for the effect of FrontVABI pressure loss on Core Driven Fan Stage(CDFS) design pressure ratio, and the calculated error of high-pressure turbine inlet total temperature is more than 40 K in mode transition condition(the transition operating condition between triple bypass mode and double bypass mode). In double bypass mode, the approximate model can better simulate the performance of FrontVABI by considering the local loss of area expansion. This method can be applied to the performance-optimized design of Front-VABI and the ACE control law design during mode transition.展开更多
Flow around a real-life underwater vehicle often happens at a high Reynolds number with flow structures at different scales from the boundary layer around a blade to that around the hull. This poses a great challenge ...Flow around a real-life underwater vehicle often happens at a high Reynolds number with flow structures at different scales from the boundary layer around a blade to that around the hull. This poses a great challenge for large-eddy simulation of an underwater vehicle aiming at resolving all relevant flow scales. In this work, we propose to model the hull with appendages using the immersed boundary method, and model the propeller using the actuator disk model without resolving the geometry of the blade. The proposed method is then applied to simulate the flow around Defense Advanced Research Projects Agency(DARPA) suboff. An overall acceptable agreement is obtained for the pressure and friction coefficients. Complex flow features are observed in the near wake of suboff. In the far wake, the core region is featured by a jet because of the actuator disk, surrounded by an annular region with velocity deficit due to the body of suboff.展开更多
In this work,the multi-fidelity(MF)simulation driven Bayesian optimization(BO)and its advanced form are proposed to optimize antennas.Firstly,the multiple objective targets and the constraints are fused into one compr...In this work,the multi-fidelity(MF)simulation driven Bayesian optimization(BO)and its advanced form are proposed to optimize antennas.Firstly,the multiple objective targets and the constraints are fused into one comprehensive objective function,which facilitates an end-to-end way for optimization.Then,to increase the efficiency of surrogate construction,we propose the MF simulation-based BO(MFBO),of which the surrogate model using MF simulation is introduced based on the theory of multi-output Gaussian process.To further use the low-fidelity(LF)simulation data,the modified MFBO(M-MFBO)is subsequently proposed.By picking out the most potential points from the LF simulation data and re-simulating them in a high-fidelity(HF)way,the M-MFBO has a possibility to obtain a better result with negligible overhead compared to the MFBO.Finally,two antennas are used to testify the proposed algorithms.It shows that the HF simulation-based BO(HFBO)outperforms the traditional algorithms,the MFBO performs more effectively than the HFBO,and sometimes a superior optimization result can be achieved by reusing the LF simulation data.展开更多
This work presents multi-fidelity multi-objective infill-sampling surrogate-assisted optimization for airfoil shape optimization.The optimization problem is posed to maximize the lift and drag coefficient ratio subjec...This work presents multi-fidelity multi-objective infill-sampling surrogate-assisted optimization for airfoil shape optimization.The optimization problem is posed to maximize the lift and drag coefficient ratio subject to airfoil geometry constraints.Computational Fluid Dynamic(CFD)and XFoil tools are used for high and low-fidelity simulations of the airfoil to find the real objective function value.A special multi-objective sub-optimization problem is proposed for multiple points infill sampling exploration to improve the surrogate model constructed.To validate and further assess the proposed methods,a conventional surrogate-assisted optimization method and an infill sampling surrogate-assisted optimization criterion are applied with multi-fidelity simulation,while their numerical performance is investigated.The results obtained show that the proposed technique is the best performer for the demonstrated airfoil shape optimization.According to this study,applying multi-fidelity with multi-objective infill sampling criteria for surrogate-assisted optimization is a powerful design tool.展开更多
In material modeling,the calculation speed using the empirical potentials is fast compared to the first principle calculations,but the results are not as accurate as of the first principle calculations.First principle...In material modeling,the calculation speed using the empirical potentials is fast compared to the first principle calculations,but the results are not as accurate as of the first principle calculations.First principle calculations are accurate but slow and very expensive to calculate.In this work,first,the H-H binding energy and H2-H2 interaction energy are calculated using the first principle calculations which can be applied to the Tersoff empirical potential.Second,the H-H parameters are estimated.After fitting H-H parameters,the mechanical properties are obtained.Finally,to integrate both the low-fidelity empirical potential data and the data from the high-fidelity firstprinciple calculations,the multi-fidelity Gaussian process regression is employed to predict the HH binding energy and the H2-H2 interaction energy.Numerical results demonstrate the accuracy of the developed empirical potentials.展开更多
This study delineates the development of the optimization framework for the preliminary design phase of Floating Offshore Wind Turbines(FOWTs),and the central challenge addressed is the optimization of the FOWT platfo...This study delineates the development of the optimization framework for the preliminary design phase of Floating Offshore Wind Turbines(FOWTs),and the central challenge addressed is the optimization of the FOWT platform dimensional parameters in relation to motion responses.Although the three-dimensional potential flow(TDPF)panel method is recognized for its precision in calculating FOWT motion responses,its computational intensity necessitates an alternative approach for efficiency.Herein,a novel application of varying fidelity frequency-domain computational strategies is introduced,which synthesizes the strip theory with the TDPF panel method to strike a balance between computational speed and accuracy.The Co-Kriging algorithm is employed to forge a surrogate model that amalgamates these computational strategies.Optimization objectives are centered on the platform’s motion response in heave and pitch directions under general sea conditions.The steel usage,the range of design variables,and geometric considerations are optimization constraints.The angle of the pontoons,the number of columns,the radius of the central column and the parameters of the mooring lines are optimization constants.This informed the structuring of a multi-objective optimization model utilizing the Non-dominated Sorting Genetic Algorithm Ⅱ(NSGA-Ⅱ)algorithm.For the case of the IEA UMaine VolturnUS-S Reference Platform,Pareto fronts are discerned based on the above framework and delineate the relationship between competing motion response objectives.The efficacy of final designs is substantiated through the time-domain calculation model,which ensures that the motion responses in extreme sea conditions are superior to those of the initial design.展开更多
For the numerical simulation of flow systems with various complex components, the traditional one-dimensional （1D） network method has its comparative advantage in time consuming and the CFD method has its absolute a...For the numerical simulation of flow systems with various complex components, the traditional one-dimensional （1D） network method has its comparative advantage in time consuming and the CFD method has its absolute advantage in the detailed flow capturing. The proper coupling of the advantages of different dimensional methods can strike balance well between time cost and accuracy and then significantly decrease the whole design cycle for the flow systems in modern machines. A novel multi-fidelity coupled simulation method with numerical zooming is developed for flow systems. This method focuses on the integration of one-, two-and three-dimensional codes for various components. Coupled iterative process for the different dimensional simulation cycles of sub-systems is performed until the concerned flow variables of the whole system achieve convergence. Numerical zooming is employed to update boundary data of components with different dimen-sionalities. Based on this method, a highly automatic, multi-discipline computing environment with integrated zooming is developed. The numerical results of Y-Junction and the air system of a jet engine are presented to verify the solution method. They indicate that this type of multi-fidelity simulationmethod can greatly improve the prediction capability for the flow systems.展开更多
The Bayesian Multi-Fidelity Surrogate(MFS)proposed by Kennedy and O’Hagan(KOH model)has been widely employed in engineering design,which builds the approximation by decomposing the high-fidelity function into a scale...The Bayesian Multi-Fidelity Surrogate(MFS)proposed by Kennedy and O’Hagan(KOH model)has been widely employed in engineering design,which builds the approximation by decomposing the high-fidelity function into a scaled low-fidelity model plus a discrepancy function.The scale factor before the low-fidelity function,ρ,plays a crucial role in the KOH model.This scale factor is always tuned by the Maximum Likelihood Estimation(MLE).However,recent studies reported that the MLE may sometimes result in MFS of bad accuracy.In this paper,we first present a detailed analysis of why MLE sometimes can lead to MFS of bad accuracy.This is because,the MLE overly emphasizes the variation of discrepancy function but ignores the function waviness when selectingρ.To address the above issue,we propose an alternative approach that choosesρby minimizing the posterior variance of the discrepancy function.Through tests on a one-dimensional function,two high-dimensional functions,and a turbine blade design problem,the proposed approach shows better accuracy than or comparable accuracy to MLE,and the proposed approach is more robust than MLE.Additionally,through a comparative test on the design optimization of a turbine endwall cooling layout,the advantage of the proposed approach is further validated.展开更多
For complex engineering problems,multi-fidelity modeling has been used to achieve efficient reliability analysis by leveraging multiple information sources.However,most methods require nested training samples to captu...For complex engineering problems,multi-fidelity modeling has been used to achieve efficient reliability analysis by leveraging multiple information sources.However,most methods require nested training samples to capture the correlation between different fidelity data,which may lead to a significant increase in low-fidelity samples.In addition,it is difficult to build accurate surrogate models because current methods do not fully consider the nonlinearity between different fidelity samples.To address these problems,a novel multi-fidelity modeling method with active learning is proposed in this paper.Firstly,a nonlinear autoregressive multi-fidelity Kriging(NAMK)model is used to build a surrogate model.To avoid introducing redundant samples in the process of NAMK model updating,a collective learning function is then developed by a combination of a U-learning function,the correlation between different fidelity samples,and the sampling cost.Furthermore,a residual model is constructed to automatically generate low-fidelity samples when high-fidelity samples are selected.The efficiency and accuracy of the proposed method are demonstrated using three numerical examples and an engineering case.展开更多
Iterative coupled methods are widely used in multi-fidelity simulation of rotating components due to the simple implementation,which iteratively eliminates the errors between the computational fluid dynamics models an...Iterative coupled methods are widely used in multi-fidelity simulation of rotating components due to the simple implementation,which iteratively eliminates the errors between the computational fluid dynamics models and approximate characteristic maps.However,the convergence and accuracy of the iterative coupled method are trapped in characteristic maps.In particular,iterative steps increase sharply as the operation point moves away from the design point.To address these problems,this paper developed an auxiliary iterative coupled method that introduces the static-pressure-auxiliary characteristic maps and modification factor of mass flow into the component-level model.The developed auxiliary method realized the direct transfer of static pressure between the high-fidelity models and the component-level model.Multi-fidelity simulations of the throttle characteristics were carried out using both the auxiliary and traditional iterative coupled methods,and the simulation results were verified using the experimental data.Additionally,the consistency between the auxiliary and traditional iterative coupled methods was confirmed.Subsequently,multi-fidelity simulations of the speed and altitude characteristics were also conducted.The auxiliary and traditional iterative coupled methods were evaluated in terms of convergence speed and accuracy.The evaluation indicated that the auxiliary iterative coupled method significantly reduces iterative steps by approximately 50%at the near-choked state.In general,the auxiliary iterative coupled method is preferred as a development of the traditional iterative coupled method in the near-choked state,and the combined auxiliary-traditional iterative coupled method provides support for successful multi-fidelity simulation in far-off-design conditions.展开更多
The rapid prediction of aerodynamic performance is critical in the conceptual and preliminary design of hypersonic vehicles. This study focused on axisymmetric body configurations commonly used in such vehicles and pr...The rapid prediction of aerodynamic performance is critical in the conceptual and preliminary design of hypersonic vehicles. This study focused on axisymmetric body configurations commonly used in such vehicles and proposed a multi-fidelity neural network (MFNN) framework to fuse aerodynamic data of varying quality. A data-driven prediction model was constructed using a pointwise modeling method based on generating lines to input geometric features into the network. The MFNN framework combined low-fidelity and high-fidelity networks, trained on aerodynamic performance data from engineering rapid computation methods and CFD, respectively, using spherically blunted cones as examples. The results showed that the MFNN effectively integrated multi-fidelity data, achieving prediction accuracy close to CFD results in most regions, with errors under 5% in key stagnation areas. The model demonstrated strong generalization capabilities for varying cone dimensions and flight conditions. Furthermore, it significantly reduced dependence on high-fidelity data, enabling efficient aerodynamic performance predictions with limited datasets. This study provides a novel methodology for rapid aerodynamic performance prediction, offering both accuracy and efficiency, and contributes to the design of hypersonic vehicles.展开更多
Low pressure ratio fans of modern civil turbofans suffer from reduced stall margin in the take-off operating line and at part-speed,requiring variable geometry devices.Variable area nozzles(VAN)are one of the investig...Low pressure ratio fans of modern civil turbofans suffer from reduced stall margin in the take-off operating line and at part-speed,requiring variable geometry devices.Variable area nozzles(VAN)are one of the investigated solutions to control engine operating conditions throughout the mission.In this paper,we present a multi-fidelity modelling approach for an ultra-high bypass ratio turbofan engine with a VAN,combining a zero-dimensional thermody-namic cycle simulator using a realistic fan map with two-and three-dimensional detailed computational fluid dynamics(CFD)simulations for internal/external flow coupling.By adopting a novel algorithm to match the cycle conditions to the CFD solutions,the propulsive performance of the turbofan is analysed in a reference aircraft mission.The numerical method captures the effect on thrust generation and nacelle drag,providing a more reliable estimation of the impact of VAN on engine operation and efficiency.Low-speed mission points are confirmed to be those that benefit the most from an enlarged fan nozzle area,with a possible improvement of 3%in terms of thrust and specific fuel consumption at take-off and approach using a 10%larger area,similarly predicted by both 2D and 3D models.A preliminary acous-tic evaluation based on semi-empirical noise models indicates a modest effect on noise emis-sions,with up to 1 dB reduction in microphone signature at the sideline for a nozzle area increased by 10%.展开更多
Acid-base dissociation constant(pK_(a)) is a key physicochemical parameter in chemical science, especially in organic synthesis and drug discovery. Current methodologies for pK_(a) prediction still suffer from limited...Acid-base dissociation constant(pK_(a)) is a key physicochemical parameter in chemical science, especially in organic synthesis and drug discovery. Current methodologies for pK_(a) prediction still suffer from limited applicability domain and lack of chemical insight. Here we present MF-SuP-pK_(a)(multi-fidelity modeling with subgraph pooling for pK_(a) prediction), a novel pK_(a) prediction model that utilizes subgraph pooling, multi-fidelity learning and data augmentation. In our model, a knowledgeaware subgraph pooling strategy was designed to capture the local and global environments around the ionization sites for micro-pK_(a) prediction. To overcome the scarcity of accurate pK_(a) data, lowfidelity data(computational pK_(a)) was used to fit the high-fidelity data(experimental pK_(a)) through transfer learning. The final MF-SuP-pK_(a) model was constructed by pre-training on the augmented ChEMBL data set and fine-tuning on the DataWarrior data set. Extensive evaluation on the DataWarrior data set and three benchmark data sets shows that MF-SuP-pK_(a) achieves superior performances to the state-of-theart pK_(a) prediction models while requires much less high-fidelity training data. Compared with Attentive FP, MF-SuP-pK_(a) achieves 23.83% and 20.12% improvement in terms of mean absolute error(MAE) on the acidic and basic sets, respectively.展开更多
In ship engineering,the prediction of vertical bending moment(VBM)and total longitudinal stress(TLS)during ship navigation is of utmost importance.In this work,we propose a new prediction paradigm,the multi-fidelity r...In ship engineering,the prediction of vertical bending moment(VBM)and total longitudinal stress(TLS)during ship navigation is of utmost importance.In this work,we propose a new prediction paradigm,the multi-fidelity regression model based on multi-fidelity data and artificial neural network(MF-ANN).Specifically,an ANN is used to learn the fundamental physical laws from low-fidelity data and construct an initial input-output model.The predicted values of this initial model are of low accuracy,and then the high-fidelity data are utilized to establish a correction model that can correct the low-fidelity prediction values.Hence,the overall accuracy of prediction can be improved significantly.The feasibility of the multi-fidelity regression model is demonstrated by predicting the VBM,and the robustness of the model is evaluated at the same time.The prediction of TLS on the deck indicates that just a small amount of high-fidelity data can make the prediction accuracy reach a high level,which further illustrates the validity of the proposed MF-ANN.展开更多
Multi-fidelity simulations incorporate computational fluid dynamics(CFD) models into a thermodynamic model,enabling the simulation of the overall performance of an entire gas turbine with high-fidelity components.Trad...Multi-fidelity simulations incorporate computational fluid dynamics(CFD) models into a thermodynamic model,enabling the simulation of the overall performance of an entire gas turbine with high-fidelity components.Traditional iterative coupled methods rely on characteristic maps,while fully coupled methods directly incorporate high-fidelity simulations.However,fully coupled methods face challenges in simulating rotating components,including weak convergence and complex implementation.To address these challenges,a fully coupled method with logarithmic transformations was developed to directly integrate high-fidelity CFD models of multiple rotating components.The developed fully coupled method was then applied to evaluate the overall performance of a KJ66 micro gas turbine across various off-design simulations.The developed fully coupled method was also compared with the traditional iterative coupled method.Furthermore,experimental data from ground tests were conducted to verify its effectiveness.The convergence history indicated that the proposed fully coupled method exhibited stable convergence,even under far-off-design simulations.The experimental verification demonstrated that the multi-fidelity simulation with the fully coupled method achieved high accuracy in off-design conditions.Further analysis revealed inherent differences in the coupling methods of CFD models between the developed fully coupled and traditional iterative coupled methods.These inherent differences provide valuable insights for reducing errors between the component-level model and CFD models in different coupling methods.The developed fully coupled method,introducing logarithmic transformations,offers more realistic support for the detailed and optimal design of high-fidelity rotating components within the overall performance platform of gas turbines.展开更多
The optimization of wings typically relies on computationally intensive high-fidelity simulations,which restrict the quick exploration of design spaces.To address this problem,this paper introduces a methodology dedic...The optimization of wings typically relies on computationally intensive high-fidelity simulations,which restrict the quick exploration of design spaces.To address this problem,this paper introduces a methodology dedicated to optimizing box wing configurations using low-fidelity data driven machine learning approach.This technique showcases its practicality through the utilization of a tailored low-fidelity machine learning technique,specifically designed for early-stage wing configuration.By employing surrogate model trained on small dataset derived from low-fidelity simulations,our method aims to predict outputs within an acceptable range.This strategy significantly mitigates computational costs and expedites the design exploration process.The methodology's validation relies on its successful application in optimizing the box wing of PARSIFAL,serving as a benchmark,while the primary focus remains on optimizing the newly designed box wing by Bionica.Applying this method to the Bionica configuration led to a notable 14%improvement in overall aerodynamic effciency.Furthermore,all the optimized results obtained from machine learning model undergo rigorous assessments through the high-fidelity RANS analysis for confirmation.This methodology introduces innovative approach that aims to streamline computational processes,potentially reducing the time and resources required compared to traditional optimization methods.展开更多
文摘To address the limitations of existing coupling methods in aero-engine system simulation,which fail to adaptively adjust iterative parameters and coupling relationships,which can result in low efficiency and in⁃stability,this study introduces a‘Dynamic Event-Driven Co-Simulation’algorithm integrated with decision tree algorithms.This algorithm separates the overall coupling relationships and the main solver from the primary mod⁃el,utilizing a dynamic event monitoring module to adaptively adjust simulation strategies,including iteration pa⁃rameters,coupling relationships,and convergence criteria.This facilitates efficient adaptive simulations of dy⁃namic events while balancing solution accuracy and computational efficiency.The research focuses on a twinshaft turbofan engine,establishing six system-level models that encompass overall performance and various sub⁃systems based on three coupling methods,along with a multidisciplinary multi-fidelity simulation framework in⁃corporating a 3D CFD nozzle model.The study tests both model exchange and coupled simulation methods under a 14 s transient acceleration and deceleration scenario.In a 100%throttle condition,a high-fidelity nozzle model is used to analyze the sensitivity of different convergence criteria on computational efficiency and accuracy.Re⁃sults indicate that the accuracy and efficiency achieved with this method are comparable to those of PROOSIS soft⁃ware(18 s and 35 s,respectively),while being 71%more efficient than Simulink software(62 s and 120 s,re⁃spectively).Furthermore,appropriately relaxing the convergence criteria for the 0D model(from 10-6 to 10-4)while enhancing those for the 3D model(from 3000 steps to 6000 steps)can effectively balance computational accuracy and efficiency.
基金supported by the National Natural Science Foundation of China(No.12102356)。
文摘The resource-intensive,high-fidelity infrared signature simulations and Radar CrossSection(RCS)calculations limit the integrated optimization of Unmanned Combat Aerial Vehicles(UCAVs)in response to escalating threats from joint detection systems.To this end,we present a sample-efficient framework to advance the optimization efficiency of UCAV's exhaust system,focusing on both the stealth characteristics evaluation and the optimization process.A novel multi-fidelity stealth assessment method,powered by multi-fidelity neural network and local perceptive fields,has been developed to fuse different fidelity information from infrared radiation signature and RCS values,respectively.Results demonstrate that the method can achieve relatively high accuracy based on a small set of high-fidelity data.Furthermore,this data fusion method is integrated into a multi-objective Bayesian optimization framework.Employing a Gaussian process regression model and the EHVI acquisition function,the framework effectively explores the stealth objective space,achieving a 15.21%hypervolume indicator increase with fewer optimization iterations compared to NSGA-Ⅱ.Results show that the optimized nozzle significantly reduces both the infrared signature and RCS compared to the baseline configuration.The proposed framework offers a practical and efficient approach for optimizing the integrated stealth performance of UCAVs.
基金funding support from the National Natural Science Foundation of China(No.52204065,No.ZX20230398)supported by a grant from the Human Resources Development Program(No.20216110100070)of the Korea Institute of Energy Technology Evaluation and Planning(KETEP)。
文摘In the realm of subsurface flow simulations,deep-learning-based surrogate models have emerged as a promising alternative to traditional simulation methods,especially in addressing complex optimization problems.However,a significant challenge lies in the necessity of numerous high-fidelity training simulations to construct these deep-learning models,which limits their application to field-scale problems.To overcome this limitation,we introduce a training procedure that leverages transfer learning with multi-fidelity training data to construct surrogate models efficiently.The procedure begins with the pre-training of the surrogate model using a relatively larger amount of data that can be efficiently generated from upscaled coarse-scale models.Subsequently,the model parameters are finetuned with a much smaller set of high-fidelity simulation data.For the cases considered in this study,this method leads to about a 75%reduction in total computational cost,in comparison with the traditional training approach,without any sacrifice of prediction accuracy.In addition,a dedicated well-control embedding model is introduced to the traditional U-Net architecture to improve the surrogate model's prediction accuracy,which is shown to be particularly effective when dealing with large-scale reservoir models under time-varying well control parameters.Comprehensive results and analyses are presented for the prediction of well rates,pressure and saturation states of a 3D synthetic reservoir system.Finally,the proposed procedure is applied to a field-scale production optimization problem.The trained surrogate model is shown to provide excellent generalization capabilities during the optimization process,in which the final optimized net-present-value is much higher than those from the training data ranges.
基金supported by the Seventh Framework Programme of China-EU Collaborative Projects
文摘Constructing high approximation accuracy surrogate model with lower computational cost has great engineering significance.In this paper,using co-Kriging method,an efficient multi-fidelity surrogate model is constructed based on two independent high and low fidelity samples.Co-Kriging method can use a greater quantity of low-fidelity information to enhance the accuracy of a surrogate of the high-fidelity model by modeling the correlation between high and low fidelity model,thus computational cost of building surrogate model can be greatly reduced.A wing-body problem is taken as an example to compare characteristics of co-Kriging multi-fidelity(CKMF)model with traditional Kriging based multi-fidelity(KMF)model.A sampling convergence of the CKMF model and the KMF model is conducted,and an appropriate sampling design is selected through the sampling convergence analysis.The results indicate that CKMF model has higher approximation accuracy with the same high-fidelity samples,and converges at less high-fidelity samples.A wing-body drag reduction optimization design using genetic algorithm is implemented.Satisfying design results are obtained,which validate the feasibility of CKMF model in engineering design.
基金funded by National Natural Science Foundation of China(Nos.51776010 and 91860205)National Science and Technology Major Project,China(No.2017-I0001-0001)。
文摘Front Variable Area Bypass Injector(Front-VABI) is a component of the Adaptive Cycle Engine(ACE) with important variable-cycle features. The performance of Front-VABI has a direct impact on the performance and stability of ACE, but the current ACE performance model uses approximate models for Front-VABI performance calculation. In this work, a multi-fidelity simulation based on a de-coupled method is developed which delivers a more accurate calculation of the Front-VABI performance based on Computational Fluid Dynamics(CFD) simulation. This simulation method proposes a form of Front-VABI characteristic and its matching calculation method between it and the ACE performance model, constructs a coupling method between the(2-D) Front-VABI model and the(0-D) ACE performance model. The result shows, when ACE works in triple bypass mode, the approximate model cannot account for the effect of FrontVABI pressure loss on Core Driven Fan Stage(CDFS) design pressure ratio, and the calculated error of high-pressure turbine inlet total temperature is more than 40 K in mode transition condition(the transition operating condition between triple bypass mode and double bypass mode). In double bypass mode, the approximate model can better simulate the performance of FrontVABI by considering the local loss of area expansion. This method can be applied to the performance-optimized design of Front-VABI and the ACE control law design during mode transition.
基金supported by the National Natural Science Foundation of China(NSFC)Basic Science Center Program for“Multiscale Problems in Nonlinear Mechanics”(No.11988102)NSFC(No.12002345)China Postdoctoral Science Foundation(No.2020M680027)。
文摘Flow around a real-life underwater vehicle often happens at a high Reynolds number with flow structures at different scales from the boundary layer around a blade to that around the hull. This poses a great challenge for large-eddy simulation of an underwater vehicle aiming at resolving all relevant flow scales. In this work, we propose to model the hull with appendages using the immersed boundary method, and model the propeller using the actuator disk model without resolving the geometry of the blade. The proposed method is then applied to simulate the flow around Defense Advanced Research Projects Agency(DARPA) suboff. An overall acceptable agreement is obtained for the pressure and friction coefficients. Complex flow features are observed in the near wake of suboff. In the far wake, the core region is featured by a jet because of the actuator disk, surrounded by an annular region with velocity deficit due to the body of suboff.
基金supported by the National Key Research and Development Program of China(2019YFB1803205)the Key Research and Development Project of Shaanxi Province(2019GY-007)+1 种基金the National Natural Science Foundation of China(61801369)the Fundamental R esearch Funds for the Central Universities(XZD012021012)。
文摘In this work,the multi-fidelity(MF)simulation driven Bayesian optimization(BO)and its advanced form are proposed to optimize antennas.Firstly,the multiple objective targets and the constraints are fused into one comprehensive objective function,which facilitates an end-to-end way for optimization.Then,to increase the efficiency of surrogate construction,we propose the MF simulation-based BO(MFBO),of which the surrogate model using MF simulation is introduced based on the theory of multi-output Gaussian process.To further use the low-fidelity(LF)simulation data,the modified MFBO(M-MFBO)is subsequently proposed.By picking out the most potential points from the LF simulation data and re-simulating them in a high-fidelity(HF)way,the M-MFBO has a possibility to obtain a better result with negligible overhead compared to the MFBO.Finally,two antennas are used to testify the proposed algorithms.It shows that the HF simulation-based BO(HFBO)outperforms the traditional algorithms,the MFBO performs more effectively than the HFBO,and sometimes a superior optimization result can be achieved by reusing the LF simulation data.
基金The authors are grateful for the support from Khon Kaen University Scholarship for ASEAN and GMS Countries’Personnel of Academic Year and the National Research Council of Thailand(N42A650549).
文摘This work presents multi-fidelity multi-objective infill-sampling surrogate-assisted optimization for airfoil shape optimization.The optimization problem is posed to maximize the lift and drag coefficient ratio subject to airfoil geometry constraints.Computational Fluid Dynamic(CFD)and XFoil tools are used for high and low-fidelity simulations of the airfoil to find the real objective function value.A special multi-objective sub-optimization problem is proposed for multiple points infill sampling exploration to improve the surrogate model constructed.To validate and further assess the proposed methods,a conventional surrogate-assisted optimization method and an infill sampling surrogate-assisted optimization criterion are applied with multi-fidelity simulation,while their numerical performance is investigated.The results obtained show that the proposed technique is the best performer for the demonstrated airfoil shape optimization.According to this study,applying multi-fidelity with multi-objective infill sampling criteria for surrogate-assisted optimization is a powerful design tool.
基金We gratefully acknowledge the support from the National Science Foundation of USA(Grants DMS-1555072 and DMS-1736364).
文摘In material modeling,the calculation speed using the empirical potentials is fast compared to the first principle calculations,but the results are not as accurate as of the first principle calculations.First principle calculations are accurate but slow and very expensive to calculate.In this work,first,the H-H binding energy and H2-H2 interaction energy are calculated using the first principle calculations which can be applied to the Tersoff empirical potential.Second,the H-H parameters are estimated.After fitting H-H parameters,the mechanical properties are obtained.Finally,to integrate both the low-fidelity empirical potential data and the data from the high-fidelity firstprinciple calculations,the multi-fidelity Gaussian process regression is employed to predict the HH binding energy and the H2-H2 interaction energy.Numerical results demonstrate the accuracy of the developed empirical potentials.
基金financially supported by the National Natural Science Foundation of China(Grant No.52371261)the Science and Technology Projects of Liaoning Province(Grant No.2023011352-JH1/110).
文摘This study delineates the development of the optimization framework for the preliminary design phase of Floating Offshore Wind Turbines(FOWTs),and the central challenge addressed is the optimization of the FOWT platform dimensional parameters in relation to motion responses.Although the three-dimensional potential flow(TDPF)panel method is recognized for its precision in calculating FOWT motion responses,its computational intensity necessitates an alternative approach for efficiency.Herein,a novel application of varying fidelity frequency-domain computational strategies is introduced,which synthesizes the strip theory with the TDPF panel method to strike a balance between computational speed and accuracy.The Co-Kriging algorithm is employed to forge a surrogate model that amalgamates these computational strategies.Optimization objectives are centered on the platform’s motion response in heave and pitch directions under general sea conditions.The steel usage,the range of design variables,and geometric considerations are optimization constraints.The angle of the pontoons,the number of columns,the radius of the central column and the parameters of the mooring lines are optimization constants.This informed the structuring of a multi-objective optimization model utilizing the Non-dominated Sorting Genetic Algorithm Ⅱ(NSGA-Ⅱ)algorithm.For the case of the IEA UMaine VolturnUS-S Reference Platform,Pareto fronts are discerned based on the above framework and delineate the relationship between competing motion response objectives.The efficacy of final designs is substantiated through the time-domain calculation model,which ensures that the motion responses in extreme sea conditions are superior to those of the initial design.
基金National Weapon Equipment Pre-research Foundation of China(0C410101110C4101)Innovation Foundation of BUAA for PhD Graduates(YWF-13-A01-15)for funding this work
文摘For the numerical simulation of flow systems with various complex components, the traditional one-dimensional （1D） network method has its comparative advantage in time consuming and the CFD method has its absolute advantage in the detailed flow capturing. The proper coupling of the advantages of different dimensional methods can strike balance well between time cost and accuracy and then significantly decrease the whole design cycle for the flow systems in modern machines. A novel multi-fidelity coupled simulation method with numerical zooming is developed for flow systems. This method focuses on the integration of one-, two-and three-dimensional codes for various components. Coupled iterative process for the different dimensional simulation cycles of sub-systems is performed until the concerned flow variables of the whole system achieve convergence. Numerical zooming is employed to update boundary data of components with different dimen-sionalities. Based on this method, a highly automatic, multi-discipline computing environment with integrated zooming is developed. The numerical results of Y-Junction and the air system of a jet engine are presented to verify the solution method. They indicate that this type of multi-fidelity simulationmethod can greatly improve the prediction capability for the flow systems.
基金the financial support from the National Science and Technology Major Project,China(No.2019-Ⅱ-0008-0028)Key Program of National Natural Science Foundation of China(No.51936008)。
文摘The Bayesian Multi-Fidelity Surrogate(MFS)proposed by Kennedy and O’Hagan(KOH model)has been widely employed in engineering design,which builds the approximation by decomposing the high-fidelity function into a scaled low-fidelity model plus a discrepancy function.The scale factor before the low-fidelity function,ρ,plays a crucial role in the KOH model.This scale factor is always tuned by the Maximum Likelihood Estimation(MLE).However,recent studies reported that the MLE may sometimes result in MFS of bad accuracy.In this paper,we first present a detailed analysis of why MLE sometimes can lead to MFS of bad accuracy.This is because,the MLE overly emphasizes the variation of discrepancy function but ignores the function waviness when selectingρ.To address the above issue,we propose an alternative approach that choosesρby minimizing the posterior variance of the discrepancy function.Through tests on a one-dimensional function,two high-dimensional functions,and a turbine blade design problem,the proposed approach shows better accuracy than or comparable accuracy to MLE,and the proposed approach is more robust than MLE.Additionally,through a comparative test on the design optimization of a turbine endwall cooling layout,the advantage of the proposed approach is further validated.
基金supported by the Major Projects of Zhejiang Provincial Natural Science Foundation of China(No.LD22E050009)the National Natural Science Foundation of China(No.51475425)the College Student’s Science and Technology Innovation Project of Zhejiang Province(No.2022R403B060),China.
文摘For complex engineering problems,multi-fidelity modeling has been used to achieve efficient reliability analysis by leveraging multiple information sources.However,most methods require nested training samples to capture the correlation between different fidelity data,which may lead to a significant increase in low-fidelity samples.In addition,it is difficult to build accurate surrogate models because current methods do not fully consider the nonlinearity between different fidelity samples.To address these problems,a novel multi-fidelity modeling method with active learning is proposed in this paper.Firstly,a nonlinear autoregressive multi-fidelity Kriging(NAMK)model is used to build a surrogate model.To avoid introducing redundant samples in the process of NAMK model updating,a collective learning function is then developed by a combination of a U-learning function,the correlation between different fidelity samples,and the sampling cost.Furthermore,a residual model is constructed to automatically generate low-fidelity samples when high-fidelity samples are selected.The efficiency and accuracy of the proposed method are demonstrated using three numerical examples and an engineering case.
基金funded by the Science and Technology Innovation Committee Foundation of Shenzhen,China(Nos.JCYJ20200109141403840 and ZDSYS20220527171405012)the National Natural Science Foundation of China(No.52106045)the Pearl River Talent Recruitment Program,China(No.2019CX01Z084)。
文摘Iterative coupled methods are widely used in multi-fidelity simulation of rotating components due to the simple implementation,which iteratively eliminates the errors between the computational fluid dynamics models and approximate characteristic maps.However,the convergence and accuracy of the iterative coupled method are trapped in characteristic maps.In particular,iterative steps increase sharply as the operation point moves away from the design point.To address these problems,this paper developed an auxiliary iterative coupled method that introduces the static-pressure-auxiliary characteristic maps and modification factor of mass flow into the component-level model.The developed auxiliary method realized the direct transfer of static pressure between the high-fidelity models and the component-level model.Multi-fidelity simulations of the throttle characteristics were carried out using both the auxiliary and traditional iterative coupled methods,and the simulation results were verified using the experimental data.Additionally,the consistency between the auxiliary and traditional iterative coupled methods was confirmed.Subsequently,multi-fidelity simulations of the speed and altitude characteristics were also conducted.The auxiliary and traditional iterative coupled methods were evaluated in terms of convergence speed and accuracy.The evaluation indicated that the auxiliary iterative coupled method significantly reduces iterative steps by approximately 50%at the near-choked state.In general,the auxiliary iterative coupled method is preferred as a development of the traditional iterative coupled method in the near-choked state,and the combined auxiliary-traditional iterative coupled method provides support for successful multi-fidelity simulation in far-off-design conditions.
文摘The rapid prediction of aerodynamic performance is critical in the conceptual and preliminary design of hypersonic vehicles. This study focused on axisymmetric body configurations commonly used in such vehicles and proposed a multi-fidelity neural network (MFNN) framework to fuse aerodynamic data of varying quality. A data-driven prediction model was constructed using a pointwise modeling method based on generating lines to input geometric features into the network. The MFNN framework combined low-fidelity and high-fidelity networks, trained on aerodynamic performance data from engineering rapid computation methods and CFD, respectively, using spherically blunted cones as examples. The results showed that the MFNN effectively integrated multi-fidelity data, achieving prediction accuracy close to CFD results in most regions, with errors under 5% in key stagnation areas. The model demonstrated strong generalization capabilities for varying cone dimensions and flight conditions. Furthermore, it significantly reduced dependence on high-fidelity data, enabling efficient aerodynamic performance predictions with limited datasets. This study provides a novel methodology for rapid aerodynamic performance prediction, offering both accuracy and efficiency, and contributes to the design of hypersonic vehicles.
文摘Low pressure ratio fans of modern civil turbofans suffer from reduced stall margin in the take-off operating line and at part-speed,requiring variable geometry devices.Variable area nozzles(VAN)are one of the investigated solutions to control engine operating conditions throughout the mission.In this paper,we present a multi-fidelity modelling approach for an ultra-high bypass ratio turbofan engine with a VAN,combining a zero-dimensional thermody-namic cycle simulator using a realistic fan map with two-and three-dimensional detailed computational fluid dynamics(CFD)simulations for internal/external flow coupling.By adopting a novel algorithm to match the cycle conditions to the CFD solutions,the propulsive performance of the turbofan is analysed in a reference aircraft mission.The numerical method captures the effect on thrust generation and nacelle drag,providing a more reliable estimation of the impact of VAN on engine operation and efficiency.Low-speed mission points are confirmed to be those that benefit the most from an enlarged fan nozzle area,with a possible improvement of 3%in terms of thrust and specific fuel consumption at take-off and approach using a 10%larger area,similarly predicted by both 2D and 3D models.A preliminary acous-tic evaluation based on semi-empirical noise models indicates a modest effect on noise emis-sions,with up to 1 dB reduction in microphone signature at the sideline for a nozzle area increased by 10%.
基金financially supported by National Key Research and Development Program of China (2021YFF1201400)National Natural Science Foundation of China (22220102001)Natural Science Foundation of Zhejiang Province (LZ19H300001, LD22H300001, China)。
文摘Acid-base dissociation constant(pK_(a)) is a key physicochemical parameter in chemical science, especially in organic synthesis and drug discovery. Current methodologies for pK_(a) prediction still suffer from limited applicability domain and lack of chemical insight. Here we present MF-SuP-pK_(a)(multi-fidelity modeling with subgraph pooling for pK_(a) prediction), a novel pK_(a) prediction model that utilizes subgraph pooling, multi-fidelity learning and data augmentation. In our model, a knowledgeaware subgraph pooling strategy was designed to capture the local and global environments around the ionization sites for micro-pK_(a) prediction. To overcome the scarcity of accurate pK_(a) data, lowfidelity data(computational pK_(a)) was used to fit the high-fidelity data(experimental pK_(a)) through transfer learning. The final MF-SuP-pK_(a) model was constructed by pre-training on the augmented ChEMBL data set and fine-tuning on the DataWarrior data set. Extensive evaluation on the DataWarrior data set and three benchmark data sets shows that MF-SuP-pK_(a) achieves superior performances to the state-of-theart pK_(a) prediction models while requires much less high-fidelity training data. Compared with Attentive FP, MF-SuP-pK_(a) achieves 23.83% and 20.12% improvement in terms of mean absolute error(MAE) on the acidic and basic sets, respectively.
基金supported by the National Key Research amd Development Program of China(Grant No.2020YFA0405700).
文摘In ship engineering,the prediction of vertical bending moment(VBM)and total longitudinal stress(TLS)during ship navigation is of utmost importance.In this work,we propose a new prediction paradigm,the multi-fidelity regression model based on multi-fidelity data and artificial neural network(MF-ANN).Specifically,an ANN is used to learn the fundamental physical laws from low-fidelity data and construct an initial input-output model.The predicted values of this initial model are of low accuracy,and then the high-fidelity data are utilized to establish a correction model that can correct the low-fidelity prediction values.Hence,the overall accuracy of prediction can be improved significantly.The feasibility of the multi-fidelity regression model is demonstrated by predicting the VBM,and the robustness of the model is evaluated at the same time.The prediction of TLS on the deck indicates that just a small amount of high-fidelity data can make the prediction accuracy reach a high level,which further illustrates the validity of the proposed MF-ANN.
基金funded by the Science and Technology Innovation Committee Foundation of Shenzhen,Grant No.JCYJ20200109141403840 and Grant No.ZDSYS20220527171405012the National Natural Science Foundation of China (NSFC),Grant No.52106045。
文摘Multi-fidelity simulations incorporate computational fluid dynamics(CFD) models into a thermodynamic model,enabling the simulation of the overall performance of an entire gas turbine with high-fidelity components.Traditional iterative coupled methods rely on characteristic maps,while fully coupled methods directly incorporate high-fidelity simulations.However,fully coupled methods face challenges in simulating rotating components,including weak convergence and complex implementation.To address these challenges,a fully coupled method with logarithmic transformations was developed to directly integrate high-fidelity CFD models of multiple rotating components.The developed fully coupled method was then applied to evaluate the overall performance of a KJ66 micro gas turbine across various off-design simulations.The developed fully coupled method was also compared with the traditional iterative coupled method.Furthermore,experimental data from ground tests were conducted to verify its effectiveness.The convergence history indicated that the proposed fully coupled method exhibited stable convergence,even under far-off-design simulations.The experimental verification demonstrated that the multi-fidelity simulation with the fully coupled method achieved high accuracy in off-design conditions.Further analysis revealed inherent differences in the coupling methods of CFD models between the developed fully coupled and traditional iterative coupled methods.These inherent differences provide valuable insights for reducing errors between the component-level model and CFD models in different coupling methods.The developed fully coupled method,introducing logarithmic transformations,offers more realistic support for the detailed and optimal design of high-fidelity rotating components within the overall performance platform of gas turbines.
基金The funding for this publication was provided by Johannes Kepler University(JKU),Linz.Special thanks to Prof.Zongmin DENG from Beihang University for his invaluable guidance,insightful feedback,and constructive criticism,which greatly enhanced the quality of this manuscript.We extend our heartfelt gratitude to the PARSIFAL team for providing the supporting materials,which inspired this study.
文摘The optimization of wings typically relies on computationally intensive high-fidelity simulations,which restrict the quick exploration of design spaces.To address this problem,this paper introduces a methodology dedicated to optimizing box wing configurations using low-fidelity data driven machine learning approach.This technique showcases its practicality through the utilization of a tailored low-fidelity machine learning technique,specifically designed for early-stage wing configuration.By employing surrogate model trained on small dataset derived from low-fidelity simulations,our method aims to predict outputs within an acceptable range.This strategy significantly mitigates computational costs and expedites the design exploration process.The methodology's validation relies on its successful application in optimizing the box wing of PARSIFAL,serving as a benchmark,while the primary focus remains on optimizing the newly designed box wing by Bionica.Applying this method to the Bionica configuration led to a notable 14%improvement in overall aerodynamic effciency.Furthermore,all the optimized results obtained from machine learning model undergo rigorous assessments through the high-fidelity RANS analysis for confirmation.This methodology introduces innovative approach that aims to streamline computational processes,potentially reducing the time and resources required compared to traditional optimization methods.
