Multiphysical optimization is particularly challenging when involving fluid–solid interactions with large deformations. While analytical approaches are commonly computational inexpensive but lack of the necessary accuracy for many applications, numerical simulations can provide higher accuracy but become very fast extremely costly. Experimental optimization approaches promise several benefits which can allow to overcome these issues in particular for application which bear complex multiphysics such as fluid–structure interactions. Here, we propose a method for an experimental optimization using genetic algorithms with a custom optimizer software directly coupled to a fully automatized experiment. Our application case is a biomimicking fish robot. The aim of the optimization is to determine the best swimming gaits for high propulsion performance in combination with low power consumption. The optimization involves genetic algorithms, more precise the NSGA-II algorithm and has been performed in still and running water. The results show a negligible impact of the investigated flow velocity. A subsequent spot analysis allows to derive some particular characteristics which leads to the recommendation to perform two different swimming gaits for cruising and for sprinting. Furthermore, we show that Exp-O techniques enable a massive reduction in the evaluation time for multiphysical optimization problems in realistic scenarios.

Particle image velocimetry (PIV) and laser-induced fluorescence (LIF) are currently state-of-the-art, non-intrusive measurement techniques that help in our understanding of heat and momentum transfer in thermal fluid flow applications. We present a simplified, integrated PIV–LIF system which simultaneously measures velocity and temperature fields in aqueous flows by means of a novel fluorescent dye (RuPhen), a low-cost continuous-wave diode laser, and a single color camera. We demonstrate that RuPhen is well-suited for this approach due to its peak absorption at 450 nm, peak emission at 605 nm, and a strong temperature-dependent emission with a sensitivity coefficient of 4%/∘C\\documentclass[12pt]{minimal} \\usepackage{amsmath} \\usepackage{wasysym} \\usepackage{amsfonts} \\usepackage{amssymb} \\usepackage{amsbsy} \\usepackage{mathrsfs} \\usepackage{upgreek} \\setlength{\\oddsidemargin}{-69pt} \\begin{document}$$4\\%/^{\\circ }\\hbox {C}$$\\end{document}. The large Stokes shift between excitation (which also includes Mie scattering of the flow tracers) and the emission facilitates the handling of the signal components in the RGB channels of the camera. To correct the recordings for laser power fluctuations, we propose a novel method that jointly employs two photoluminescence signals. We provide a spectral characterization of the dye at different temperatures and discuss the choice of each component for our measurement system. We demonstrate the potential of our approach by two experimental test cases that focus on thermally driven and turbulent flow regimes.

The measurement of the transport of scalar quantities within flows is oftentimes laborious, difficult or even unfeasible. On the other hand, velocity measurement techniques are very advanced and give high-resolution, high-fidelity experimental data. Hence, we explore the capabilities of a deep learning model to predict the scalar quantity, in our case temperature, from measured velocity data. Our method is purely data-driven and based on the u-net architecture and, therefore, well-suited for planar experimental data. We demonstrate the applicability of the u-net on experimental temperature and velocity data, measured in large aspect ratio Rayleigh–Bénard convection at Pr=7.1\\documentclass[12pt]{minimal} \\usepackage{amsmath} \\usepackage{wasysym} \\usepackage{amsfonts} \\usepackage{amssymb} \\usepackage{amsbsy} \\usepackage{mathrsfs} \\usepackage{upgreek} \\setlength{\\oddsidemargin}{-69pt} \\begin{document}$$\ extrm{Pr} =7.1$$\\end{document} and Ra=2×105,4×105,7×105\\documentclass[12pt]{minimal} \\usepackage{amsmath} \\usepackage{wasysym} \\usepackage{amsfonts} \\usepackage{amssymb} \\usepackage{amsbsy} \\usepackage{mathrsfs} \\usepackage{upgreek} \\setlength{\\oddsidemargin}{-69pt} \\begin{document}$$\ extrm{Ra} =2\ imes 10^5,4\ imes 10^5,7\ imes 10^5$$\\end{document}. We conduct a hyper-parameter optimization and ablation study to ensure appropriate training convergence and test different architectural variations for the u-net. We test two application scenarios that are of interest to experimentalists. One, in which the u-net is trained with data of the same experimental run and one in which the u-net is trained on data of different Ra\\documentclass[12pt]{minimal} \\usepackage{amsmath} \\usepackage{wasysym} \\usepackage{amsfonts} \\usepackage{amssymb} \\usepackage{amsbsy} \\usepackage{mathrsfs} \\usepackage{upgreek} \\setlength{\\oddsidemargin}{-69pt} \\begin{document}$$\ extrm{Ra}$$\\end{document}. Our analysis shows that the u-net can predict temperature fields similar to the measurement data and preserves typical spatial structure sizes. Moreover, the analysis of the heat transfer associated with the temperature showed good agreement when the u-net is trained with data of the same experimental run. The relative difference between measured and reconstructed local heat transfer of the system characterized by the Nusselt number Nu\\documentclass[12pt]{minimal} \\usepackage{amsmath} \\usepackage{wasysym} \\usepackage{amsfonts} \\usepackage{amssymb} \\usepackage{amsbsy} \\usepackage{mathrsfs} \\usepackage{upgreek} \\setlength{\\oddsidemargin}{-69pt} \\begin{document}$$\ extrm{Nu}$$\\end{document} is between 0.3 and 14.1% depending on Ra\\documentclass[12pt]{minimal} \\usepackage{amsmath} \\usepackage{wasysym} \\usepackage{amsfonts} \\usepackage{amssymb} \\usepackage{amsbsy} \\usepackage{mathrsfs} \\usepackage{upgreek} \\setlength{\\oddsidemargin}{-69pt} \\begin{document}$$\ extrm{Ra}$$\\end{document}. We conclude that deep learning has the potential to supplement measurements and can partially alleviate the expense of additional measurement of the scalar quantity.

We explore the strengths and limitations of using a standard Michelson interferometer to sample line-of-sight-averaged temperature in water via two experimental setups: slow-varying temperature in static fluid and fast temperature variations in convective flow. The high precision of our measurements (a few mK) is enabled by the fast response time and high sensitivity of the interferometer to minute changes in the refractive index of water caused by temperature variations. These features allow us to detect the signature of fine fluid dynamical patterns in convective flow in a fully non-intrusive manner. For example, we are able to observe an asymmetry in the rising thermal plume (i.e., an asynchronous arrival of two counter-rotating vortices at the measurement location), which is not possible to resolve with more traditional (and invasive) techniques, such as RTD (Resistance Temperature Detector) sensors. These findings, and the overall reliability of our method, are further corroborated by means of Particle Image Velocimetry and Large Eddy Simulations. While this method presents inherent limitations (mainly stemming from the line-of-sight-averaged nature of its results), its non-intrusiveness and robustness, along with the ability to readily yield real-time, highly accurate measurements, render this technique very attractive for a wide range of applications in experimental fluid dynamics.

Full-scale drag penalty predictions of flows over rough walls require surface roughness characterisation from laboratory experiments or numerical simulations. In either approach, it is necessary to determine the so-called equivalent sand-grain roughness height (k_s). There are several steps involved in determining this aerodynamic roughness lengthscale, but its procedure typically includes a combination of measurement of wall-shear stress (tau _w) using direct or indirect methods as well as analysis of velocity profiles. Indirect methods usually rely on assumptions made about flow and its scaling including the validity of universal outer-layer similarity. However, the implications of the underlying assumptions involved in full-scale drag prediction are unclear. In this work, we carry out wind tunnel measurements over a realistic rough surface (from a fouled ship-hull) to evaluate the impact of different methods with an emphasis on using the outer-layer similarity hypothesis for full-scale drag predictions. Wall-shear stress is measured using an in-house floating-element drag balance (DB), and velocity profiles are obtained using particle image velocimetry (PIV), allowing the evaluation of k_s, and the associated wake parameters through several methods. The aerodynamic roughness parameters hence obtained are used for full-scale drag penalty calculations. It is observed that the predicted drag penalty can vary by over 15% among the different methods highlighting the care that should be taken when employing such methods.