Smith C, Doherty JJ, Jin Y (2014) Convergence Based Prediction Surrogates for High-lift CFD Optimization,
In this paper we analyze the convergence profiles of computational fluid dynamics calculations for a multi-element high-lift system, revealing insights into the flow physics of the system. The dependencies of CL on differing numbers of flow solver iterations are also investigated.
Using a surrogate model to evaluate the fitness of candidate solutions in an evolutionary algorithm can significantly reduce the overall computational cost of optimization tasks. Therefore, a hybrid multi-objective evolutionary algorithm, that trains and optimizes the structure of a recurrent neural network ensemble, is introduced as a surrogate, for the long-term prediction of the high-lift systems computational fluid dynamic convergence data. The intermediate data is used for training the networks and results show that the trends of the design space can be predicted using a quarter of the flow solver iterations.
Mehdizadeh Gavgani A, Bingham T, Sorniotti A, Doherty J, Cavallino C, Fracchia M (2015) A Parallel Hybrid Electric Drivetrain Layout with Torque-Fill Capability, SAE International Journal of Passenger Cars - Mechanical Systems 8 (2) pp. 767-778
Copyright © 2015 SAE International.This paper discusses the torque-fill capability of a novel hybrid electric drivetrain for a high-performance passenger car, originally equipped with a dual-clutch transmission system, driven by an internal combustion engine. The paper presents the simulation models of the two drivetrains, including examples of experimental validation during upshifts. An important functionality of the electric motor drive within the novel drivetrain is to provide torque-fill during gearshifts when the vehicle is engine-driven. A gearshift performance indicator is introduced in the paper, and the two drivetrain layouts are assessed in terms of gearshift quality performance for a range of maneuvers.
Tan X, Wang J, Xu Y, Curran R, Raghunathan S, Gore D, Doherty J (2008) Risk analysis in lifecycle costing of aluminium, 12th AIAA/ISSMO Multidisciplinary Analysis and Optimization Conference, MAO
Aluminium is extensively applied in aerospace industry, however, uncertainties lie in all activities of its production. In this paper, risk assessment for cost estimates in aluminium lifecycle is investigated. Through the in-house module, COSTLIPS (COSTing Life-cycle In material ProcesseS), probabilities of typical processes for aluminium production are estimated. Approach of probability distribution functions and three-point estimates is used to estimate statistical parameters of the costs for different processes. Example results show that the COSTLIPS can be used for costing of aluminium production processes including risk analysis. This would be useful for material selection and evaluation in aircraft design and manufacture in term of cost estimation. Copyright © 2008 by the American Institute of Aeronautics and Astronautics, Inc.
Xu Y, Wang J, Tan X, Curran R, Raghunathan S, Doherty J, Gore D (2008) A generic life cycle cost modeling approach for aircraft system, Collaborative Product and Service Life Cycle Management for a Sustainable World - Proceedings of the 15th ISPE International Conference on Concurrent Engineering, CE 2008 pp. 251-258
Life cycle cost (LCC) is truly representative to the total cost of an aircraft through its life cycle. It is usually used for estimating the cost-effectiveness of an aircraft design. For enabling LCC estimation at early stage, A LCC model have been being developed for aircraft wing under the umbrella of Integrated Wing Advanced Technology Validation Programme in United Kingdom. Object-oriented and hierarchical approaches are used for LCC modelling. The cost estimation is based on bottom-up approach. The developed LCC model is generic, and can be customized and applied for estimating the costs of other aircraft systems. © 2008 Springer-Verlag London Limited.
Smith C, Doherty J, Jin Y (2013) Recurrent neural network ensembles for convergence prediction in surrogate-assisted evolutionary optimization, Proceedings of the 2013 IEEE Symposium on Computational Intelligence in Dynamic and Uncertain Environments, CIDUE 2013 - 2013 IEEE Symposium Series on Computational Intelligence, SSCI 2013 pp. 9-16
Evaluating the fitness of candidate solutions in evolutionary algorithms can be computationally expensive when the fitness is determined using an iterative numerical process. This paper illustrates how an ensemble of Recurrent Neural Networks can be used as a robust surrogate to predict converged Computational Fluid Dynamics data from unconverged data. The training of the individual neural networks is controlled and a variance range is used to determine if the surrogates have been adequately trained to predict diverse and accurate solutions. Heterogeneous ensemble members are used due to the limited data available and results show that for certain parameters, predictions can be made to within 5% of the converged data's final output, using approximately 40% of the iterations needed for convergence. The implications of the method and results presented are that it is possible to use ensembles of Recurrent Neural Networks to provide accurate fitness predictions for an evolutionary algorithm and that they could be used to reduce the time needed to achieve optimal designs based on time-consuming Computational Fluid Dynamics simulations. © 2013 IEEE.
Smith C, Doherty J, Jin Y (2014) Multi-objective evolutionary recurrent neural network ensemble for prediction of computational fluid dynamic simulations, Proceedings of the 2014 IEEE Congress on Evolutionary Computation, CEC 2014 pp. 2609-2616
© 2014 IEEE.Using a surrogate model to evaluate the expensive fitness of candidate solutions in an evolutionary algorithm can significantly reduce the overall computational cost of optimization tasks. In this paper we present a recurrent neural network ensemble that is used as a surrogate for the long-term prediction of computational fluid dynamic simulations. A hybrid multi-objective evolutionary algorithm that trains and optimizes the structure of the recurrent neural networks is introduced. Selection and combination of individual prediction models in the Pareto set of solutions is used to create the ensemble of predictors. Five selection methods are tested on six data sets and the accuracy of the ensembles is compared to the converged computational fluid dynamic data, as well as to the delta change between two flow conditions. Intermediate computational fluid dynamic data is used for training and the method presented can produce accurate and stable results using a third of the intermediate data needed for convergence.
The work presented here is an on-going part of the UK Department for Business, Enterprise & Regulatory Reform (DBERR)/Industry funded Integrated Wing programme. In particular within the Configuration Optimisation and Integration sub-task, work funded by DBERR and QinetiQ is focused on assessment of a multi-disciplinary design optimisation (MDO) approach for supporting conceptual design. This approach can potentially satisfy the requirements for general applicability and accuracy through use of high-fidelity, physicsbased simulation, but it is essential to demonstrate that the approach represents a practical method which can be automated and fast enough to meet the turn-around times required within conceptual design. In particular it is necessary to identify the level of detail which must be incorporated into MDO to provide a balance between accuracy, generality and speed. In order to help identify this necessary level of detail, a baseline MDO capability has been established which will be incrementally enhanced through improvements to the modelling of specific technologies and systems. The study will also provide an indication of how an aircraft configuration, designed using MDO, changes as a result of the impact of integrating different technologies and systems. This latter outcome supports a key objective for the Integrated Wing programme by potentially helping to identify those technologies and systems which can lead to a step change in performance for future civil aircraft. This paper presents progress towards these goals, including a description of the MDO approach being followed and the integration of additional technologies and systems into the process. © 2007 by QinetiQ Ltd.
Tan X, Xu Y, Wang J, Curran R, Raghunathan S, Gore D, Doherty J (2008) Life cycle assessment of aluminium for engineering application, 8th AIAA Aviation Technology, Integration and Operations (ATIO) Conference
Material of aluminium alloys is a typical material applied in Aerospace Engineering. In order to conduct life cycle assessment of aluminium, a module based on Microsoft Excel has been developed. Through this module, COSTing Life-cycle In material ProcesseS (COSTLIPS) can be made. This paper presents the methodology to the development of the COSTLIPS and its results of estimation based on typical processes for manufacturing aluminium components. Through the COSTLIPS, it can be estimated not only processing costs, but also materials input and output, energy consumed, and wastes emitted. COSTLIPS could be a useful tool to help understand economic, energy and environmental impacts when an aluminium component/part used in engineering. Copyright © 2008 by the American Institute of Aeronautics and Astronautics, Inc. All rights reserved.
Chew JW, Doherty JW, Gillan M, Hills NJ (2006) practical applications of automated design and optimisation techniques using CFD,
Application of design optimisation techniques using CFD to problems for turbomachinery internal air systems and motorsport will be described and discussed. Developments in these areas build on earlier work on turbomachinery blade and aircraft applications, and present new challenges. Specific examples include turbine cooling air pre-swirl nozzles, turbine rim sealing, and full track optimisation of a Champ Car. Focussing on these specific examples, issues such as choice of optimisation method, automation of mesh generation, geometry parameterisation, solution convergence, model accuracy and method robustness will be considered.
Xu Y, Wang J, Tan X, Early J, Curran R, Raghunathan S, Doherty J, Gore D (2008) Life cycle cost modeling for aircraft wing using object-oriented systems engineering approach, 46th AIAA Aerospace Sciences Meeting and Exhibit
Life Cycle Cost (LCC) is an important issue for aircraft design, and can be used for the development of trade studies. An Object-Oriented System Engineering (OOSE) approach has been developed to establish a LCC model architecture, and a bottom-up approach is adopted to conduct cost estimation. The established LCC model can represent the impact of technology, and can be supplemented by increasing the fidelity functional sub-models. The developed model is generic, and as such can be easily transferable for applications in other disciplines. Copyright © 2008 by the American Institute of Aeronautics and Astronautics, Inc.
During the early stages of aircraft design it is typically the case that potential solutions are down-selected based upon a relatively low level of supporting information. It is also often the case that major airframe characteristics are effectively fixed by the time serious consideration is given to the integration of aircraft systems. As a result, the integration of systems may be significantly more complex and may lead to greater performance penalties than would be the case if these integration requirements were better addressed during early design. The evolution of new capabilities for supporting early aircraft design, based upon Multi-disciplinary Design Optimisation (MDO) toolsets, can potentially enable more complex requirements to be directly addressed in early design. In order to assess the potential opportunities resulting from the use of MDO, consideration has been given to the requirement for improved modelling of fuel systems within early civil aircraft wing design. Details of the MDO approach used are presented. Enhancements to the MDO method, associated with the integration of fuel tanks within the wing subject to uncontained engine rotor failure (UERF) considerations, are described. The results of an initial parametric variation study are presented. The reported work has been carried out as part of the Integrated Wing collaborative programme, which is part funded by the UK Department for Business, Enterprise & Regulatory Reform (DBERR). Copyright © 2009 by QinetiQ Ltd.
The work presented here is an on-going part of the UK Department for Business, Enterprise & Regulatory Reform (DBERR)/Industry funded Integrated Wing programme. In particular within the Configuration Optimisation and Integration sub-task, work funded by DBERR and QinetiQ is focused on assessment of a multi-disciplinary design optimisation (MDO) approach for supporting conceptual design. This approach can potentially satisfy the requirements for general applicability and accuracy through use of high-fidelity, physics-based simulation, but it is essential to demonstrate that the approach represents a practical method which can be automated and fast enough to meet the turn-around times required within conceptual design. In particular it is necessary to identify the level of detail which must be incorporated into MDO to provide a balance between accuracy, generality and speed. In order to help identify this necessary level of detail, a baseline MDO capability has been established which will be incrementally enhanced through improvements to the modelling of specific technologies and systems. The study will also provide an indication of how an aircraft configuration, designed using MDO, changes as a result of the impact of integrating different technologies and systems. This latter outcome supports a key objective for the Integrated Wing programme by potentially helping to identify those technologies and systems which can lead to a step change in performance for future civil aircraft. This paper presents progress towards these goals, including a description of the MDO approach being followed and the integration of additional technologies and systems into the process.
Tan X, Wang J, Xu Y, Curran R, Raghunathan S, Gore D, Doherty J (2008) Cost-efficient materials in aerospace: Composite vs aluminium, Collaborative Product and Service Life Cycle Management for a Sustainable World - Proceedings of the 15th ISPE International Conference on Concurrent Engineering, CE 2008 pp. 259-266
Aluminium alloys are a series of traditional materials applied in aerospace industry. Composite materials have been being used to replace some aluminium alloys in airframe structures. To select a cost-efficient material in aerospace design, a framework for systematic assessment of composites and aluminium is therefore suggested. Through evaluation of 3Ps which represent prices, properties and processing of material, life cycle assessment of aluminium and composite could be made. By comparison of and trade-off the listed baselines of 3Ps, an economic, rational, available and useful material can be logically selected. © 2008 Springer-Verlag London Limited.
Doherty JJ, McParlin SC (2004) Generic process for air vehicle concept design and assessment, 42nd AIAA Aerospace Sciences Meeting and Exhibit Online Proceedings pp. 11418-11427
In recent years, the UK Ministry of Defence (MOD) has funded development, by QinetiQ and its predecessor organisations, of processes and tools to assess the performance of air vehicles, with the objective of maintaining status as an intelligent customer for a variety of air vehicle types. During this period, Operational Requirements have been evolving, requiring increased flexibility and the capability to produce accurate performance data for novel air vehicle concepts, including those which are not adequately represented by existing semi-empirical methods and databases. In order to explain the assessment process that has been developed, an example manned aircraft application is described. The component parts of the assessment process, and the underlying techniques and technologies are also described. Finally, indications are given of possible future directions.
Tan X, Wang J, Xu Y, Raghunathan S, Gore D, Doherty J (2008) Costing of aluminium for life cycle, 46th AIAA Aerospace Sciences Meeting and Exhibit
Aluminium alloys are a key series of metallic materials used in aerospace engineering. Life cycle of aluminium consists of a series of independent product processes including, for example, bauxite mining, bauxite refining, smelting, casting, forming, performing, and disposal. In order to understand the life cycle cost of aluminium, an in-house module for costing of life cycle in material processes has been developed. The object oriented costing approach is suggested and adopted. Based on survey data and engineering analysis of the processing, cost estimating relationships are established. Through the module, costs for various processes of aluminium production can be estimated. Example results show that the module could be a useful tool for costing life cycle of aluminium, or for estimating costs for an aluminium component with an individual process. Copyright © 2008 by the American Institute of Aeronautics and Astronautics, Inc.
An initial investigation of an optimisation based approach for design across a continuous range of operating conditions is presented. The objective for this 'operations based optimisation' approach is to avoid the need to choose critical design point conditions and associated weighting factors by tackling the overall operational performance instead. The approach integrates numerical optimisation, response surface modelling, CFD and operational simulation. An optimisation test bed involving the aerodynamic optimisation of a Champ Car rear wing assembly for reduced lap time using track simulation has been developed to assess the new optimisation approach. Details of the operations based optimisation approach and the Champ Car test bed are reported. Results generated using the new approach are presented and the wider potential of the approach for aerospace applications is discussed.
Doherty JJ, Dean SRH, Ellsmore P, Eldridge A (2008) A Multi-Fidelity Approach for Supporting Early Aircraft Design Decisions, In: Curran R, Chou S-Y, Trappey AJC (eds.), Collaborative Product and Service Life Cycle Management for a Sustainable World pp. 267-279 Springer
The QinetiQ Aerospace Consultancy group has been actively developing and applying process automation and optimisation capabilities in support of air vehicle assessment and design for over 20 years. These capabilities have evolved greatly during this timeframe from their initial origins as research activities, into mature capabilities for underpinning decision making in both civil and military air vehicle projects. In parallel the same generic approaches have also found usage in weapons, maritime and motorsport design. In recent years effort has focussed on enhancing a number of different, but complementary, capabilities at QinetiQ, each of which have different advantages and disadvantages, but which together better address the needs of air vehicle assessment and design. These capabilities are linked by a common requirement to assess widely differing characteristics concurrently, in order to model the consequences of design decisions, such as technology and system choices, in terms of the overall impact on an air vehicle project. This paper describes these tools in the context of their use within the Integrated Wing project.
Function evaluations of many real-world optimization
problems are time or resource consuming, posing a serious
challenge to the application of evolutionary algorithms to solve
these problems. To address this challenge, the research on
surrogate-assisted evolutionary algorithms has attracted increasing
attention from both academia and industry over the past
decades. However, most existing surrogate-assisted evolutionary
algorithms either still require thousands of expensive function
evaluations to obtain acceptable solutions, or are only applied to
very low-dimensional problems. In this paper, a novel surrogateassisted
particle swarm optimization inspired from committeebased
active learning is proposed. In the proposed algorithm,
a global model management strategy inspired from committeebased
active learning is developed, which searches for the best
and most uncertain solutions according to a surrogate ensemble
using a particle swarm optimization algorithm and evaluates
these solutions using the expensive objective function. In addition,
a local surrogate model is built around the best solution obtained
so far. Then a particle swarm optimization algorithm searches
on the local surrogate to find its optimum and evaluates it. The
evolutionary search using the global model management strategy
switches to the local search once no further improvement can be
observed, and vice versa. This iterative search process continues
until the computational budget is exhausted. Experimental results
comparing the proposed algorithm with a few state-of-the-art
surrogate-assisted evolutionary algorithms on both benchmark
problems up to 30 decision variables as well as an airfoil design
problem demonstrate that the proposed algorithm is able to
achieve better or competitive solutions with a limited budget
of hundreds of exact function evaluations.
A number of transonic airfoils, designed using differing approaches, are evaluated over a
wide range of operating conditions, using a tool for generating aerodynamic performance
maps. Details of key performance boundaries are also extracted, including drag divergence
and separation onset. The aerodynamic performance maps and boundaries, which are based
upon extensive use of a rapid 2D CFD tool, are first demonstrated on an existing airfoil, for
which the design condition is known and for which experimental data is available.
Aerodynamic maps are then presented for a series of airfoils which are designed using a
sonic plateau, inverse design approach. Further maps are presented for airfoils designed
using single-point and multi-point optimization. The impact of the alternative design
approaches is studied, using the performance maps and the resulting characteristics of the
performance boundaries. In particular, the trade-off between drag divergence and the onset
of separation, combined with viscous and wave drag development, is presented. The study
provides some insights into the challenge of achieving a well posed optimization formulation
for transonic airfoil design.
Many real-world optimization problems involve
computationally intensive numerical simulations to accurately
evaluate the quality of solutions. Usually the fidelity of the
simulations can be controlled using certain parameters and there
is a trade-off between simulation fidelity and computational cost,
i.e., the higher the fidelity, the more complex the simulation
will be. To reduce the computational time in simulation-driven
optimization, it is a common practice to use multiple fidelity levels
in search for the optimal solution. So far, not much work has been
done in evolutionary optimization that considers multiple fidelity
levels in fitness evaluations. In this work, we aim to develop
test suites that are able to capture some important characteristics
in real-world multi-fidelity optimization, thereby offering
a useful benchmark for developing evolutionary algorithms for
multi-fidelity optimization. To demonstrate the usefulness of the
proposed test suite, three strategies for adapting the fidelity
level of the test problems during optimization are suggested
and embedded in a particle swarm optimization algorithm. Our
simulation results indicate that the use of changing fidelity is able
to enhance the performance and reduce the computational cost
of the particle swarm optimization, which is desired in solving
expensive optimization problems.
In solving many real-world optimization problems,
neither mathematical functions nor numerical simulations are
available for evaluating the quality of candidate solutions. Instead,
surrogate models must be built based on historical data
to approximate the objective functions and no new data will be
available during the optimization process. Such problems are
known as offline data-driven optimization problems. Since the
surrogate models solely depend on the given historical data, the
optimization algorithm is able to search only in a very limited
decision space during offline data-driven optimization. This paper
proposes a new offline data-driven evolutionary algorithm to
make the full use of the offline data to guide the search. To this
end, a surrogate management strategy based on ensemble learning
techniques developed in machine learning is adopted, which
builds a large number of surrogate models before optimization
and adaptively selects a small yet diverse subset of them during
the optimization to achieve the best local approximation accuracy
and reduce the computational complexity. Our experimental
results on the benchmark problems and a transonic airfoil design
example show that the proposed algorithm is able to handle
offline data-driven optimization problems with up to 100 decision
For multi-scenario airfoil shape optimization problems,
an evaluation of a single airfoil is based on its full-scenario
drag landscape. To obtain the full-scenario drag landscape, a
large number of computational fluid dynamic simulations for
different operating conditions must be conducted. Since a single
computational fluid dynamic simulation is often time-consuming,
evaluations for multi-scenario airfoil shape optimization will
be computationally highly intensive. Although surrogate-assisted
evolutionary algorithms have been widely applied to expensive
optimization problems, existing surrogate-assisted evolutionary
algorithms cannot be directly applied to multi-scenario airfoil
shape optimization. Instead of using surrogate models to directly
approximate the multi-scenario evaluations, we employ a hierarchical
surrogate model consisting of a K-nearest neighbors classifier
and a Kriging model to approximate the full-scenario drag
landscape for each candidate design during the optimization.
Then, the fitness of the candidate design is evaluated based on
the approximated drag landscape to reduce the computational
cost. The proposed hierarchical surrogate model is embedded in
the covariance matrix adaptation evolution strategy and applied
to the RAE2822 airfoil design problem. Our experimental results
show that the proposed algorithm is able to obtain an airfoil
design with limited computational cost that perform well in
different operating conditions.
This paper presents a series of transonic airfoils, designed using differing optimization approaches, which are evaluated over a wide range of operating conditions using global aerodynamic performance maps. Global drag rise boundaries, which are identified, modelled and directly optimized during design, include drag divergence and onset of wave drag. The AIAA ADODG Case 2 airfoil optimization case is used to compare the results of the new global performance design approach with conventional multi-point optimization. The impact of alternative design formulations is presented in terms of both global performance maps and selected drag rise characteristics around the Case 2 design condition. In particular, the trade-off between drag divergence and the preceding onset of wave drag is discussed. The new approach addresses the issue of early excessive drag creep, which is typically encountered for optimization focused on a narrow range of operating conditions. The study provides some further insights into how a well posed optimization formulation for transonic airfoil design can potentially be established.
Research covers propulsion installation characteristics for novel air-vehicle configurations in the class of small
missile/UAVs. A modular experimental model has been developed featuring a fully integrated tail mounted inline
boundary layer ingesting (BLI) Electric Ducted Fan (EDF) propulsion system. The research aims to
investigate the aerodynamic characteristics and gain understanding of the optimisation of a fully integrated
propulsion system. A combination of experimental and validated CFD methods will be used to develop suitable
performance metrics and optimisation approaches for in-line BLI EDFs, thus enabling the generation of a toolset
to be applied in the future design and implementation of this class of propulsion system. Progress in both
experimental design and computational analysis will be presented, including initial analysis of the overall flow
physics, focussing on the upstream and downstream effect of the body boundary layer, the condition of the flow
ingested into the fan and an assessment of the performance of the overall integrated EDF.
As conventional aircraft designs approach their limits in terms of efficiency and emissions, a drastic change to the architecture of conventional platforms is required if the environmental targets of the next several decades are to be met. Boundary Layer Ingestion is one of industry?s most promising answers to the challenges of the future, identifying a potential step-change in performance in more integrated propulsion and airframe systems. This paper investigates the behaviour of a boundary layer ingesting solution of a closely embedded wing-electric ducted fan design, with focus on the implications of the aerodynamic coupling on the individual performance of both the aerodynamic and propulsive elements as well as on the assessment of the reliability of a low order panel code method. Wind tunnel testing was undertaken to understand the flow physics at different combinations of airframe and propulsor operating conditions; in addition, part of the data used for the experimental validation of a panel method model for predicting the upstream inlet flow conditions. It was found that there were clear local and extended upstream effects of the propulsor on the performance of the aerodynamic surface, resulting from the different combinations of suction strength and nacelle blockage. Similar trends were observed in the numerical code predictions, and identified limitations of the methodology in defining the experimental boundary conditions of the propulsor to be imposed in CFD. The study of the response of the propulsor to varying inlet boundary conditions, created by varying wing angles of attack was also carried out, however, small changes in flow velocity combined with measurement errors of the current system, prevented any solid conclusions being drawn about the impact of distorted inlet flow on propulsor performance.
Automatic optimisers can play a vital role in the design and development of engineering systems and
processes. However, a lack of available data to guide the search can result in the global optimum solution
never being found. Surrogate models can be used to address this lack of data and allow more of the
design space to be explored, as well as provide an overall computational saving.
In this thesis I have developed two novel long-term prediction methods that investigate the use of ensembles
of surrogates to perform predictions of aerodynamic data. The models are built using intermediate
computational fluid dynamic convergence data. The first method relies on a gradient based learning
algorithm to optimise the base learners and the second utilises a hybrid multi-objective evolutionary algorithm.
Different selection schemes are investigated to improve the prediction performance and the
accuracy of the ensembles are compared to the converged data, as well as to the delta change between
Three challenging real world aerodynamic data sets have been used to test the developed algorithms
and insights into aerodynamic performance has been gained through analysis of the computational fluid
dynamic convergence histories. The trends of the design space can be maintained, as well as achieving
suitable overall prediction accuracy. Selecting a subset improves ensemble performance, but no selection
method is superior to any others. The hybrid multi-objective evolutionary algorithm approach is also
tested on two standard time series prediction tasks and the results presented are competitive with others
reported in the literature.
In addition, a novel technique that improves a parameter based surrogates learning through the transfer
of additional information is also investigated to address the lack of data. Transfer learning has an initial
impact on the learning rate of the surrogate, but negative transfer is observed with increasing numbers
of epochs. Using the data available for the low dimensional problems, it is shown that the convergence
prediction results are comparable to those from the parameter based surrogate. Therefore, the convergence
prediction method could be used as a surrogate and form part of an aerodynamic optimisation task.
However, there are a number of open questions that need to be addressed, including what is the best use
of the surrogate during the search?