This article explains a data visualization of large-scale unsteady flow simulation using high-order unstructured methods in aeronautical engineering. First, a data structure of the flux reconstruction (FR) method is presented focusing on a difference from the conventional finite-volume method which has been used for steady flow simulations such as the Reynolds averaged numerical simulation. Then, we introduce a showcase for a large-scale direct numerical simulation around a low-pressure turbine blade. The simulation was performed by an open-source flow solver PyFR, and a visualization of unsteady vortex field is presented with a brief introduction of an in-situ visualization with PyFR. Finally, the outlook and possible issues are discussed for the in-situ visualization of unsteady flow simulation.

  In recent years, MR(mixed reality) technology is attracting attention as a means of transmitting various types of information. In particular, with the development of high-performance MR devices, various companies are moving to incorporate MR technology as a new information transmission device. The Visualization Team of JAXA's Supercomputer Division has been focusing on the technology of displaying 3D numerical simulation results as a new visualization technology that goes beyond the conventional 2D monitor display. It is practical for users to examine the results of their own simulations. In this paper, we introduce the problems that must be solved in order to provide that technology to users. In addition, as a technological goal to be pursued in the near future, we also introduce our approach to the technology to directly display the high-resolution 3D numerical simulation results computed with supercomputers to the MR device.

In this paper, the visualizations of tens to hundred billion grid-points large-scale high-fidelity LES (large-eddy simulation) data in aeronautics computed on the supercomputer Fugaku are introduced. The visualizations of (1) flow phenomena around complex full-aircraft geometry and (2) input-output analysis of complex flows are discussed in how we visualize the large-scale data and understand the physics of complex flows.

In the turbulent transition of hypersonic boundary layer, the secondary mode fluctuation at a frequency (several hundred kHz) higher than the primary mode (TS wave) plays a dominant role. In recent years, the secondary mode fluctuation grows by interfering with the low frequency fluctuation has been observed in hypersonic wind tunnel tests, but the conventional linear stability analysis cannot explain this, and it is necessary to establish a new theory considering the nonlinear effect. The author’s research group aims to capture the growth process of secondary modes that interfere with low-frequency modes by direct numerical simulation (DNS) and to obtain the data necessary to construct a nonlinear theory. In this paper, we attempted to generate turbulent transitions by adding a low-frequency temperature disturbance of 10 kHz to the boundary layer through an energy source term. As a result, we successfully captured the process of turbulent transitions in which high-frequency secondary modes different from the added frequency are excited by DNS.

  This article introduces visualization examples at Aerodynamics Laboratory, Yokohama National University: 1) Shockwave detection and visualization using edge-detection coined as CEDRIC (Canny-Edge-Detection / Rankine-hugonIot-Conditions unified shock sensor), and 2) vortex visualization by velocity vectors on a semi-transparent cross-section.

The present paper introduces the visualization methodology of large-scale numerical analysis results of the atomization and the evaporation of fuel in the aircraft engine combustor. An analysis solver based on the VoF method is developed, and large-scale numerical analyses are conducted using 31.2 billion computational cells for the atomization analysis and 2.2 billion cells for the evaporation analysis. The analysis data is enormous on both analyses, and the costs of the visualization are over the CFD analysis cost in some cases. We achieve a smooth visualization environment with such large-scale data using in-situ visualization and conversion of three-dimensional data to iso-surface data.

Combustion chamber of liquid-propellant rocket engine is developed mainly based on tests using subscale chamber with single and multiple injectors and full-scale engine. Understanding the combustion phenomena only from the test is difficult, and the knowledge obtained in the subscale test is not always applicable to the full-scale chamber. Development of numerical simulation technique to predict the onset of combustion instability and evaluate wall heat flux is required. In this article, the compressible Large-Eddy Simulation was applied to the subscale chambers to validate the numerical method. Mechanism of combustion instability was also discussed. Then, present challenge for the analysis of the full-scale rocket chamber by the compressible Large-Eddy Simulation was also presented.

Atmospheric turbulence is an important issue for both the operational efficiency and safety of aircraft. In particular, turbulence around airports makes it difficult to maintain regularity in landing and return operations and affects safety and comfort during takeoff and landing.

Computational fluid dynamics is necessary to understand the spatio-temporal characteristics of turbulence in detail. The visualization expected by researchers is different from that expected by pilots and flight operators in terms of making use of it for operational decisions. This paper presents a visualization of turbulent airflow around an airport using a 3D immersive VR visualization system to help pilots and flight operators understand the extent of turbulence. We introduce the visualization results of terrain-induced turbulence at Shonai Airport and hangar wake on the runway at Haneda Airport.

  Cyber-physical systems (CPS) can be realized in various ways depending on the target; however, the cyber and physical should be integrated statistically to consider their uncertainties by a Bayesian approach, when we consider computer-aided engineering (CAE) simulations as a cyber representative and corresponding measurement data as a physical counterpart. Data assimilation, which realizes Bayesian estimation using numerical simulations and measurement data, is a method of statistically estimating the uncertain factors of the simulation model (initial and boundary conditions, model parameters, etc.) with observed data. With this, the data assimilation is expected to promote cyber-physical synchronization. In this paper, we introduce an example of cyber-physical synchronization through video images.

A simple method for identifying vortex structures is proposed for the analysis of unsteady flow fields. The proposed method is applied to the flow field around NACA0015 airfoil to extract and visualize vortex structures. As the result of applying to the flow field, it is shown that the results are comparable to the isosurfaces of the second invariant of the velocity gradient tensor. In addition, by visualizing multiple time series data at once using only those with dominant spanwise vorticity components, it is shown that the advection and location of vortex breakdown can be easily identified.

The transonic buffet is a phenomenon of aerodynamic instability due to the oscillation of a shock wave; its generating mechanism has not been fully solved. This paper introduces an analysis of the transonic buffet with an Evolutionary Rule-based Learning (ERL) approach. ERL can extract generalized knowledge as a set of rules while solving a given machine learning problem. In the presented analysis, ERL is used to extract essential features likely provoking turbulent separated flow. Further, the impulse response analysis is applied to extracted features to investigate how those features affect the temporal propagations of pressures towards the shock-wave oscillation.

  In this paper, we introduce a novel design-informatics approach to extract and visualize human experts’ knowledge about airplane design. Using the dataset of candidate designs for Silent Super-Sonic Technology Demonstrator (S3TD) developed by Japan Aerospace Exploration Agency (JAXA), we try to extract and visualize the expert’s chosen design among 58 candidates in a systematic manner. Using Mapper, which is one of the representative methods in topological data analysis, we visualize a four-dimensional trade-off relationship as a directed graph. Our method found the expert’s choice as a knee point of the Pareto front, which implies that our approach can systematically select candidate designs that human experts may prefer.

Since fluid phenomena have complex spatio-temporal behaviors, it is still not easy to find physically important features in the fluid data obtained from experiments and numerical analyses. Dynamic mode decomposition (DMD), which is introduced in this paper, is one of the data analysis methods that have been actively studied in recent years, and can extract spatial and temporal features inherent in fluid data. DMD has the advantage of being applicable to a wide variety of data because it does not assume any governing equations. In this paper, some examples of the research on dynamic mode decomposition are briefly introduced.

  It is known that "generic" surface flows are Morse-Smale flows, which are generalizations of gradient flows, and that incompressible surface flows are not Morse-Smale. When we analyze fluid phenomena on 3 dimensional spaces, we use flows on slices which can be obtained by projections on the slices. In this case, such flows on slices need not be neither Morse-Smale nor incompressible in general, and so we needed to extend the existing theories to describe the topologies of "generic" surface flows. Therefore we introduced a topological invariant and its representation, called COT representation, on a large class of surface flows, which contains all Morse-Smale flows and generic incompressible spherical flows. In this issue, we explain the representations of flows which is easy to use even for non-experts and is suitable to be implemented by computers and to describe not only the topologies of such flows and also the transitions among them. Moreover, the representations can be applied by phenomena which can be regarded as flows.