Experiences at CTR with Numerical Simulation Databases Parviz Moin CTR History • Rogallo & Moin (1984) Annual Review of Fluid Mechanics (page 133): “…we have already reached the point of being able to generate more information than we are able to digest. One can imagine in the near future a researcher at a graphics terminal with access to computed turbulent flow fields of high resolution… The computer can answer any question about the fields it holds, and the researcher can devote his time to the really difficult effort of finding the right questions to ask……” “The use of stored simulation results places fewer constraints on the questions that can be answered, and allows rapid interactive display of results… Furthermore the flow field data can be shared by other researchers who do not have the computer power required to generate the fields, but do have enough power to probe them.” “The development of hardware and software tools for interactive probing of simulation results, the availability of computed flow fields… will ultimately determine the degree to which simulation enhances our understanding and ability to control turbulence.” The Center for Turbulence Research • Founded in 1987 • Archived direct numerical simulation databases for general use • Established the credibility and the physical realism of the simulation fields through numerous publications • The DNS data were declared trustworthy by well known experimentalists • The Summer Programs were established • The theme of all Summer Programs has been “Studying Turbulence using Numerical Simulation Databases” CTR • Well known experimentalists and theoreticians were invited and participated in the early Summer Programs (Adrian, Bradshaw, Blackwelder, Hunt, Jiménez, Hussain, Ho, Huerre, Alfredsson, Johansson, Landahl, …) • 9 Summer Programs (Next one in 2004) • Recent Summer Programs have limited use of Ames’ archived data. Many participants use their own data or perform calculations during the program. • CTR’s databases have also been available outside of the Summer Programs. However, most requests were for processed data (e.g. Reynolds stress budgets) From Experience at CTR • The availability of raw (unprocessed) data is important. • The use of simulation data should be facilitated for non-DNSers, such as experimentalists. • For each dataset, provide basic processing tools such as FFT’s and utilities for derivatives (See Dresselhaus (CTR Briefs, 1995) for the development of the “Tensoral Language”). Also an example script would be very helpful. Questions/Issues • Should the person who generated the DNS data be expected to be a co-author in any publication that uses his data? No. Reference to the first publication that used the data should be sufficient. • When should the data be released? When validated and after first publication by the author. • Concern for being upstaged in making a discovery by someone using your data. (This was an easy thing to do, and I was planning to do it myself.) In the big picture, it won’t matter. Besides most easy problems were addressed when DNS was novel. Questions/Issues • Misuse of data. (Example: KMM velocity fields are not divergence-free!) Not a real issue. The quality of data speaks for itself. However, the price of this is some handholding.