CFD-enabled design optimisation of industrial flows–theory and practise

The use of and interest in Computational Fluid Dynamics (CFD) and their variants have been growing at an ever-increasing rate over the last five decades and applied in numerous industries. The future development of many complex products and processes will be based on a systematic, model-based process where computational design optimisation methods will be a key enabling technology.

Optimisation based on physics-based computer simulations of complex systems is increasingly being used to solve a wide variety of challenging design problems in science and industry. This approach is now very well established for structural design problems and is now used routinely to minimise the weight of automotive components or design composite wings for aircraft. Although there have been comparatively few studies which have used CFD to optimise complex flow problems, a survey of the recent literature reveals that interest in CFD-enabled design optimisation is now growing rapidly. This special edition has been designed to present a concise overview of the key theoretical developments and industrial applications that are driving the rapid growth in this field.

There have been several major achievements that have enabled the time needed to optimise complex flows using CFD simulations to be reduced significantly. The recent review by Haftka et al. (2016) noted the rapid progress in reducing computational times for both gradient-free and gradient-based optimisation methods. The former have been shown to be very effective for up to 100 design variables, whereas for larger design problems gradient-based methods, powered by rapid advances in adjoint methods, have solved problems where the number of design variables is in the 1000s. Other key improvements have been made in adaptive sampling methods, which can provide an appropriate balance between exploration and exploitation, and multi-fidelity modelling which enables most of the computational work to be done on cheaper, lower fidelity models. Both of these enable the number of expensive, high fidelity simulations used in the optimisation process to be kept to a minimum. Significant progress has also been made in multi-disciplinary design optimisation (MDO) methods which coordinate simulations of the individual disciplines affecting a design (e.g. fluid mechanics, structural mechanics, heat transfer, …) toward a system design that is optimal as a whole, taking into account the competing objectives (Aissa et al., 2016).

There are several exciting research directions that will enable CFD-enabled design optimisation methods to have even greater impact in the future. The ACARE Beyond 2020 Vision (European Commission, 2019), for example, predicts that exploiting CFD simulations within effective MDO methods will be a key enabling technology for the future development of environment-friendly aircraft. For these and other safety-critical applications (in for example the nuclear industry), there will be increasing demands for the development of robust CFD-enabled optimisation methods that can ensure that product and/or process performance does not degrade significantly due to unavoidable variations in manufacturing tolerances, operating conditions, etc. It is also likely that the growing interest in using Machine Learning within CFD analyses, for example to tune parameters in turbulence models, and the increasing trend of combining physics-based and data-driven flow simulations will widen the both the power and scope of CFD-enabled design optimisation methods in the future.

The papers in this special edition have been carefully selected to describe important theoretical advances that will drive the future growth of CFD-enabled optimisation of industrial flows. These include adjoint methods for CAD-based shape optimisation, MDO methods for industrial workflows, adaptive sampling methods for multi-fidelity CFD modelling, Machine Learning for surrogate model calibration and methods for minimising the effect of numerical noise on the optimisation process. These concepts are illustrated through important and novel applications in the aerospace, automotive, chemical processing and naval engineering sectors.

The papers were internationally refereed, each of them being evaluated by two independent reviewers. Therefore, we must acknowledge those colleagues whose anonymous work is greatly appreciated. We also thank the authors who have worked hard to prepare the full versions of their papers. Finally, we are greatly indebted to Professor Wagdi G. Habashi, Editor-in-Chief of IJCFD, who kindly approved the publication of the Special Issue and who was also very supportive along the entire review process.