Data-Driven Surrogate-Assisted Evolutionary Fluid Dynamic Optimisation
Lead Research Organisation:
University of Surrey
Department Name: Computing Science
Abstract
Abstracts are not currently available in GtR for all funded research. This is normally because the abstract was not required at the time of proposal submission, but may be because it included sensitive information such as personal details.
People |
ORCID iD |
Yaochu Jin (Principal Investigator) | |
John Doherty (Co-Investigator) |
Publications
Cheng R
(2015)
A Multiobjective Evolutionary Algorithm Using Gaussian Process-Based Inverse Modeling
in IEEE Transactions on Evolutionary Computation
Cheng R
(2018)
Model-based evolutionary algorithms: a short survey
in Complex & Intelligent Systems
Du W
(2019)
High-Dimensional Robust Multi-Objective Optimization for Order Scheduling: A Decision Variable Classification Approach
in IEEE Transactions on Industrial Informatics
He C
(2019)
Accelerating Large-Scale Multiobjective Optimization via Problem Reformulation
in IEEE Transactions on Evolutionary Computation
He C
(2019)
A repository of real-world datasets for data-driven evolutionary multiobjective optimization
in Complex & Intelligent Systems
Hua Y
(2019)
A Clustering-Based Adaptive Evolutionary Algorithm for Multiobjective Optimization With Irregular Pareto Fronts.
in IEEE transactions on cybernetics
Huang Y
(2020)
Decision-making and multi-objectivization for cost sensitive robust optimization over time
in Knowledge-Based Systems
Jin Y
(2019)
Data-Driven Evolutionary Optimization: An Overview and Case Studies
in IEEE Transactions on Evolutionary Computation
Liu Y
(2017)
A Many-Objective Evolutionary Algorithm Using A One-by-One Selection Strategy.
in IEEE transactions on cybernetics
Description | Our main findings so far are: 1) New surrogate-assisted evolutionary algorithms for both on-line and offline data-driven optimization of high-dimensional problems. This enables the use of surrogate-assisted evolutionary algorithms to more real-world complex problems that have a large number of decision variables. 2) We have developed a surrogate-management strategy that can adapt the needed fidelity of the surrogates that can reduce computation time while maintaining correct convergence. In addition, we developed efficient algorithms for multi-fidelity and multi-scenario optimization to deal with more complex optimization problems. 3) Methods for offline evolutionary optimization based on transfer learning have been developed and verified on real-world problems 4) Methods for Bayesian evolutionary optimization for high-dimensional and many-objective problems have been developed by proposing a multi-objective infill criterion, using an heterogeneous ensemble and a dropout deep neural network. 5) Methodologies for transfer optimization, i.e., to speed up the evolutionary optimization problem at hand by transferring knowledge from previous optimization tasks. |
Exploitation Route | The new data-driven evolutionary optimization problems make it possible to solve complex real-world problems. The algorithms have been applied to optimization of a ventilation system in a large agricultural tractor by Valtra, a company in Finland, and are currently being applied to optimization of vehicle dynamics and hybrid electric vehicle controller by Honda in Germany. These algorithms have also been used by Airbus UK for high-lift wing systems design. Most recently, new research projects with Bosch and Honda have been set up to apply findings from this project and develop new methods for vehicle dynamic optimization and electric drive optimization. Some ideas will also be applied to optimization of deep neural networks for secure and robust learning. In 2018, we were awarded a Royal Society Exchange program continuing research on Bayesian evolutionary optimization. |
Sectors | Aerospace, Defence and Marine,Digital/Communication/Information Technologies (including Software),Energy,Manufacturing, including Industrial Biotechology,Security and Diplomacy,Transport |
Description | Out of the research outcome of of this award, i.e., new efficient surrogate-assisted evolutionary optimization algorithms, we have established research projects with Honda in Germany to optimize vehicle dynamics, hybrid electric vehicle controller and vehicle crash tests. The developed methods have also been used in optimization of a ventilation system at Valtra and will be used in electric drive design at Bosch. We are now extending methodologies from this project to a new research area, namely evolutionary neural architecture search, which has similar challenges as in evolutionary optimization of aerodynamic systems. |
Sector | Energy,Transport |
Impact Types | Societal,Economic |
Description | Decision Support for Complex Multiobjective Optimization Problems |
Amount | € 866,322 (EUR) |
Organisation | Finnish Funding Agency for Innovation |
Sector | Public |
Country | Finland |
Start | 01/2015 |
End | 12/2017 |
Description | Many-objective optimization for vehicle dynamics |
Amount | £13,200 (GBP) |
Organisation | Honda R&D Europe Deutschland |
Sector | Private |
Country | Germany |
Start | 04/2018 |
End | 03/2021 |
Description | Multi-source side information fusion assisted Bayesian optimization |
Amount | £12,000 (GBP) |
Funding ID | IEC\NSFC\170279 |
Organisation | The Royal Society |
Sector | Charity/Non Profit |
Country | United Kingdom |
Start | 04/2018 |
End | 03/2020 |
Description | Surrogate based runtime difference mitigation in asynchronous multi-disciplinary search tasks |
Amount | £148,429 (GBP) |
Organisation | Honda Research Institute Europe GmbH |
Sector | Private |
Country | Germany |
Start | 01/2019 |
End | 12/2022 |
Title | PlatEMO |
Description | We have developed a software tool called PlatEMO, which is a user-friendly software in Matlab, collecting over 100 evolutionary algorithms for solving single- and multi-objective optimization problems. Since it is publicized in 2017, it has attracted increasing users from academia and industry from all over the world. As of today, the tool has been reported in over 700 papers. Details of the tool are reported in the following paper: Ye Tian, Ran Cheng, Xingyi Zhang, and Yaochu Jin. PlatEMO: A MATLAB platform for evolutionary multi-objective optimization. IEEE Computational Intelligence Magazine, 12(4): 73-87, 2017 which won the 2019 Outstanding Paper Award. |
Type Of Material | Improvements to research infrastructure |
Year Produced | 2017 |
Provided To Others? | Yes |
Impact | This tool makes it much easier for non-expert researchers to develop and test new data-driven evolutionary optimization algorithms, and for industrial practitioners to design and optimize products. As we know, the tool is daily used by Honda and other industry. |
URL | https://github.com/BIMK/PlatEMO |
Description | Industrial Partner |
Organisation | Aerospace Technology Institute |
Country | United Kingdom |
Sector | Charity/Non Profit |
PI Contribution | The research project aims to develop computationally efficient and effective evolutionary algorithms for optimization based on CFD simulations, which is important for design aircraft, or parts of the aircraft. One problem we are working on is the design of high-lift wing systems, which we hope will contribute to design at Airbus and QinetiQ associated with Aerospace Technology Institute (ATI). |
Collaborator Contribution | ATI is going to provide us CFD data and some software tools for our research on developing surrogate-assisted evolutionary algorithms. |
Impact | Two papers have been published (resulted from a previous project): 1. Christopher Smith, John Doherty, Yaochu Jin: Multi-objective evolutionary recurrent neural network ensemble for prediction of computational fluid dynamic simulations. IEEE Congress on Evolutionary Computation 2014: 2609-2616 2. Christopher Smith, John Doherty, and Yaochu Jin. Recurrent neural network ensembles for convergence prediction in surrogate-assisted evolutionary optimisation. IEEE Symposium on Computational Intelligence in Dynamic and Uncertain Environments, part of the IEEE Symposium Series on Computational Intelligence, Singapore, 16-19 April 2013 |
Start Year | 2011 |
Description | Joint project |
Organisation | University of Exeter |
Country | United Kingdom |
Sector | Academic/University |
PI Contribution | This is a joint project. The team in Exeter focuses on Bayesian learning based optimization while we are working on strategies for surrogate management for evolutionary optimization based on computational fluid dynamics simulations. |
Collaborator Contribution | We can benefit the Bayesian Optimization based evolutionary algorithms to be incorporated in our surrogate-assisted evolutionary optimization. We are also building up the same platform for structure optimization based on CFD simulations. |
Impact | We are co-organizing a workshop on surrogate-assisted evolutionary optimization at the 2016 Genetic and Evolutionary Computation Conference to be held in Denver, USA. |
Start Year | 2012 |
Description | 2018 International Workshop on Data-driven Optimization of Complex Systems and Applications |
Form Of Engagement Activity | Participation in an activity, workshop or similar |
Part Of Official Scheme? | No |
Geographic Reach | International |
Primary Audience | Professional Practitioners |
Results and Impact | The workshop aims to promote data-driven evolutionary optimization and their industrial applications. Audience come from both industry and academia. At both workshops, Prof Y Jin has presented the main research findings of the project as a Keynote speaker.. |
Year(s) Of Engagement Activity | 2017,2018 |
Description | GECCO Workshop on Surrogate-Assisted Evolutionary Optimization |
Form Of Engagement Activity | Participation in an activity, workshop or similar |
Part Of Official Scheme? | No |
Geographic Reach | International |
Primary Audience | Professional Practitioners |
Results and Impact | The workshop was organised with the 2016 Genetic and Evolutionary Computation Conference held in Denver, USA. Two invited presentations, four submitted contributions were presented at the Workshop. About 30 researchers attended the workshop. |
Year(s) Of Engagement Activity | 2016 |
URL | http://gecco-2016.sigevo.org/index.html/Workshops |