Numerical Computations Database for Supersonic and Hypersonic Fluid Simulations

ONE OF THE MORE DIFFICULT PROBLEMS IN AEROSPACE ENGINEERING today is the accurate prediction of high speed fluid flow. Its numerical simulation requires the integration of knowledge coming from several different scientific disciplines. These disciplines include, in no particular order, fluid mechanics (complex flow analysis), basic physics and chemistry (models), experimentation (analysis and validation), numerical analysis (algorithms) and computer science (high-performance computing and visualization). High-powered supercomputers capable of solving large two and three dimensional high speed flow problems are no longer restricted to those in large research centers. Today's multi-megaflop computer/workstation is inexpensive and found in the smallest research lab. What isn't typically found in small research labs, however, is the necessary diversity of technical expertise to tackle such complex flow problems. One group may be skilled in deriving efficient numerical algorithms but not have the personnel or software needed to generate three space dimensional grids. Another group may be particularly competent in physical modeling and yet may be somewhat isolated and not have access to high quality experimental data.

Arguably, one of the most significant developments of this decade is the advance of the World Wide Web. This new technology makes possible the dissemination of knowledge in a manner unimaginable only a few years ago. The goal of this project is to take advantage of this new technology to make readily available state-of-the-art information and high quality data concerning high speed fluid flow to the world community.

History of the Database
An international cooperative effort was initiated in the 1980s to study various aspects of high speed fluid flows. Short Courses on Hypersonics were held in Paris (France), Colorado Springs (USA) and Aachen (Germany), and in the early 1990s, three sessions (Parts I, II and III) of a scientific workshop organized by INRIA and GAMNI took place in Antibes, France. Following Part II of the Antibes series, it was deemed necessary to preserve the essential part of the scientific knowledge acquired during the HERMES vehicle R&D Program. It appeared to be of major interest both to collect the data of the principal experimental and computational contributions to the workshop series and to develop an electronic archive that would stimulate future scientific investigations in the domain of high-velocity flow analysis and its related numerical methodologies.

In 1995, a sequel workshop was held for the first time in the United States. This workshop, organized by and held at the University of Houston, provided an incentive to develop the concept of the high speed flow electronic database. The result was the Web-based and interactive Houston High Speed Flow Database (HHSFD) which opened in the summer of 1997. The HHSFD incorporates much of the data collected during the earlier European workshops as well as new test cases and data collected at the November 1997 workshop, organized by CIRA, and held in Naples, Italy. New test cases with supporting computational and experimental data were to have been incorporated into the HHSFD following the November 1998 workshop titled "First Eastern-Western High Speed Flow Fields Conference and Workshop" to be held in Kyoto, Japan.

Organization and Content
The HHSFD database is organized in a hierarchical fashion based on Unix directories. The top-most directory is named Workshops. Under this are directories associated to specific high speed flow workshops: 1993Antibes, 1994Noordwijk, 1995Houston, and 1997Napoli. The database was intentionally designed to make the insertion of new workshops and/or the modification of existing workshops a straightforward matter. Under a given workshop directory are directories associated with workshop test case problems and below these are directories associated to their test case subproblems. (For example, a given test case may have several far field conditions or modellizations.) At the test case subproblem level, there is a bifurcation of directory trees associated to computational contributions and experimental contributions-both branches, however, share the same organizational structure. The lowest directories in the Workshops tree are those that contain the computational or experimental test case results from workshop contributors. These directories follow the naming convention Surname.Initial. Figure 1 depicts a typical Workshops branch.

Fig. 1. HHSFD Workshops directory structure.

Special files are found at several levels of the Workshops hierarchy that contain relevant database information. They are appended by either .info (specially formatted informational files) or by .ps (Postscript files).

User's Guide
The HHSFD's Home Page URL is: http://hhsfd.math.uh.edu. Here, you are asked to login. Enter your full email address and password in the indicated text fields and press submit. If you do not have an HHSFD supplied password, enter guest. Your username (email address) and password are both case sensitive. When logging in, your Web browser's Javascript³ capability is recorded and will remain static during the entire HHSFD session. If you are running a version of Netscape that supports JavaScript 1.1, check that Javascript is enabled under Network Preferences before linking to the HHSFD Home Page. While Javascript is not necessary to exploit all database capabilities, its use produces a superior interface and is suggested. If you are using Internet Explorer as your Web browser, Javascript will not be used. It is important to check that your computer has a version of winzip or gzip installed and your browser is set up to handle content encoding x-gzip with one of these helpers. Also check that a Postscript viewer such as gsview or ghostview is installed and your browser is set up to handle content type application/postscript. All Postscript files are compressed with gzip.

Fig. 2. HHSFD Web schematic.

Figure 2 depicts the overall layout of Web links within the HHSFD. Most are self-explanatory and warrant no further comment here. Two areas are worth some additional explanation, however. The first is under the heading Miscellaneous Features on the Index Page and is titled The HHSFD database search facility. Every contributor to the HHSFD has supplied a file contrib.info which contains certain biographical information as well as information specific to the contribution. The search database feature allows the database user to search for information contained within this file by category. As an example, to search for all contributions from "Marco Marini" made at the "1997 Napoli Workshop," select category name and workshop 1997Napoli, type Marco Marini in the text field and press OK. The search page will reload with links supplied to all relevant Contributor Pages. A second page worth further mention is the Contributor Page itself. From this page it is possible to download the contributor's methodological Postscript article. For computational contributions, it is possible to download and/or visualize the supplied computed solution contained in the contributor's solution.hhdb file. Solution files can be quite large, up to 100 megabytes each. Images generated by selecting a flow variable to view are generated by database software on the fly and are completed within a second or two for even the largest three dimensional dataset. These images are transferred in the jpeg format and are typically on the order of 30 kilobytes.

Control and Operation
The master database resides on an IBM N40 RS6000-based laptop machine running IBM's AIX variant of the Unix operating system. This machine was chosen because it is portable, has sufficient processing capability, runs an extremely stable operating system and provides access to SCSI peripherals without requiring additional devices. The machine runs the Apache Web server. Apache is the most popular Web server in use today and was chosen for its efficiency and portability.

The user's interface to the database (i.e., nearly all of the viewable Web pages) are generated dynamically based on information contained within the database itself. This is made possible by a programming language developed specifically for this project to facilitate the task of building the database. HERL (the Houston Extraction and Replacement Language) is a simple but fully featured language which is output-centric in that all text recognized neither as a programming construct nor a comment is passed on verbatim as output. This makes it especially well suited for embedding into Web pages as nearly every legal HTML document is also a legal HERL program. HERL constructs are inserted into the document between a pair of backwards slashes '\.' Variables in HERL always appear between a pair of dollar signs '$' and are expanded anywhere they appear.

HERL is generally run through the Common Gateway Interface, where the Web server provides it with certain arguments and environment settings. The source document is then read, any HERL constructs within are interpreted and the result is sent as output to the client's browser. Most pages within the database begin by reading standard files which set common variables and validate the user's security settings. Index pages are formed by taking files in various directories and generating HTML to display appropriate descriptions and icons for them. The result is a dynamic display of the contents of the database that does not require manual editing of any index pages.

References
¹Hypersonic Flows for Reentry Problems. Eds. J.-A. Desideri, R. Glowinski and J. Periaux, Springer, 1990.
²S. Dreilinger. "CVS Version Control for Web Development," URL: http://interactivate.com/public/cvswebsites/howto-cvs-websites.pdf.
³D. Flanagan. JavaScript: The Definitive Guide. O'Reilly and Assoc., 1996.
⁴J. MacDonald, P. Hilfinger and L. Semenzato, PRCS: The Project Revision Control System, 8th Int'l Symp. on System Configuration, URL: http://www.xcf.berkeley.edu/~jmacd/prcs.html. (Submitted.)
⁵R. Sanders and J. Tibbitts. "An Internet Accessible Database for High Speed Fluid Flow Code Validation," Proc. of ECCOMAS 1998, John Wiley & Sons. (To appear.)

Publications
Amundson, N. R., E. Morano, and R. Sanders. "Techniques for the Numerical Solution of Steady Reaction-Diffusion Systems Employing Stefan-Maxwell Diffusion," East-West J. of Numerical Mathematics (1998).
Houston High Speed Flow Database. URL: http://hhsfd.math.uh.edu.
Morano, E. and R. Sanders. The Carbuncle Phenomena: On Upwind Schemes in Multidimensions. Les Comptes Rendus of the French Academy of Sciences, 1997. 339-46.
Sanders, R. and J. Tibbitts. "An Internet Accessible Database for High Speed Fluid Flow Code Validation," Proc. of the ECCOMAS (European Community on Computational Methods in Applied Sciences) 1998 Conf., John Wiley & Sons.
Sanders, R., E. Morano, and M.-C. Druguet. "Multidimensional Dissipation for Upwind Schemes: Stability and Applications to Gas Dynamics," J. of Computational Physics. (Accepted for publication.)
Presentations
First Europe-U. S. High Speed Flow Field Database Workshop, opening lecture, Napoli, Italy, 1997.
NASA Langley Research Center, CFD Seminar, Langley, VA, 1997.
AIAA Snowmass Conf., special 3 hour presentation, Snowmass, Co, 1997.
Funding
"High Speed Flow Database." NASA Grant NAG9896, ($30,000).

Contents
ISSO -- Institute for Space Systems Operations
1997-1998 Annual Report