SGI Altix Used to Help Make Space Shuttle Flight Safe

With NASA now projecting a mid-July Space Shuttle return to flight, Lockheed Martin Space Systems-Michoud Operations continues its impact analysis of the external fuel tank using the SGI Altix computing system and SGI InfiniteStorage purchased from Silicon Graphics in 2004.

Michoud, which designs and assembles the Space Shuttle external fuel tank for NASA, made the purchase to complete impact analysis simulations of foam, ice and other debris and to model/analyze the design of the shuttle’s external tank. Michoud chose SGI high- performance compute technology because SGI benchmark tests demonstrated to Michoud engineers that they could run the three- million-plus calculations required in time to deliver their findings by NASA deadlines.

Shuttle return-to-flight preparations are occurring throughout NASA facilities and space centers. Michoud, in New Orleans, is a government-owned, contractor-operated component of the NASA Marshall Space Flight Center in Huntsville, Ala. Engineers at Lockheed Martin- operated Michoud facility are working closely with the Ice Formation Team at NASA Kennedy Space Center as well as Glenn Research Center, which delivers the computational fluid dynamic (CFD) models Michoud uses for impact analysis.

As part of the effort to ensure the safest possible return to flight, the structural dynamics group at Michoud was charged with debris impact analysis on the shuttle’s external fuel tank.

Potential debris elements include foam and ice and the analysis is location-specific, such as foam off the inner tank flange or foam off the forward ogive (pronounced oh-jive) – which is the aerodynamicallyshaped forward end of the tank.

Other elements include solid rocket booster (SRB) exhaust products, SRB ablator (material used to transfer heat away from high heat-erosion areas) and even butcher paper, which is put up around the motors to keep out moisture. The elements range in size from small masses of .0001 pounds all the way up to some ice particles as large as .09 pounds. Possible trajectories of each one of these pieces of debris – where they could go and where they could hit – result in millions of combinations.

In preparation to run the estimated 3 million-plus variations of possible debris impact analysis, Michoud’s Information Technology Systems (ITS) department considered a number of hardware and software vendors to meet the staggering computational demands. Based on the criteria set by the NASA agency-wide team, SGI and LS-DYNA software (a generalpurpose transient dynamic finite element program) from Livermore Software Technology Corp., became the obvious choice, as Jimmy Blevins, structural dynamics engineer at Lockheed Martin Space Systems, Michoud explained.

“Right now, we can’t simulate real world behavior of an external tank in something like a wind tunnel test,” Blevins said. “It’s absolutely impossible to do, and there is no way they were going to let me shoot at a real tank filled with liquid propellant. We had to rely upon high-performance computers to run simulation code that gives us the behavior of how a piece of foam would come off a tank.

“We had to come up with an analytical method to correlate that testing effort with what would actually happen to the tank in flight. And the only way to get there was with LS-DYNA, and the only way to deal with LS-DYNA on the scale that we required was to go with the SGI system.”

Blevins, who almost single-handedly runs the impact analysis on the SGI Altix system, added, “SGI ran some benchmarks for us with LS- DYNA. The results were compelling. A job that would have taken eight days to run on our current system took a little over four hours to run on the Altix server.

The model sizes we’re running now are much bigger than the benchmark tests. I’m running jobs that are about 30 hours each, which would have taken months, if they had even been possible to run at all. I don’t believe these LS-DYNA calculations, at the detail level we need, would ever have run on another machine. With the SGI Altix system, we basically compressed two years worth of work into about three months.”

Lockheed Martin Spaces SystemsMichoud, a long-time SGI customer, completed installation of a SGI Altix 3000 server with 64GB RAM with the Linux operating environment running on 32 Intel Itanium 2 processors, as well as a 4TB SGI InfiniteStorage TP9100 solution, last July. ITS systems administrators Leslie Jennings and Bob Meibaum were originally charged with defining the user requirements and selecting the appropriate hardware.

“What the SGI Altix server offered was shared memory as opposed to a cluster of PCs,” said Meibaum. “In a cluster, each PC has a certain amount of memory on it and then you have to communicate between the PCs, but you can’t see the memory: you’re communicating between the processors on the PCs. With the Altix system, the memory is accessible by any of the processors. Any of the 32 processors on this Altix system can access any of the 64 gigabytes of memory. Shared memory makes for a much more flexible and faster machine.”

“We reviewed benchmarks and talked to software vendors and they said that the SGI Altix server ran various applications really well, and LS-DYNA ran very, very well on Altix system,” Jennings said. “Another reason we selected Altix server is that I really like the idea of having one company behind the hardware and the Linux software. A lot of times when you buy a Linux box, you have your hardware from here and your software from there and then system support is hard to get. But I know I can always count on support from SGI, and that meant a lot to me as a system administrator. Now that we see the power of the Altix system in our impact analysis, we are considering the purchase of a similarly configured Altix system for future work at Michoud.”

At the same time as the SGI Altix system acquisition, Michoud, for a separate external tank re-design effort, purchased 11 Silicon Graphics visual workstations each with IGBRAM.

The workstations are being used to remodel portions of the original external tank design database, which was limited to only wireframe design capabilities, into the CATIA CAD-CAM application from Dassault Systemes.

CATIA enables Michoud engineers to make a solid from the wireframe model, shade it, and do impact and volume analysis. In a sense, CATIA, powered by SGI visualization workstations, is used to draw a complex “skin” around the wireframe models, which was used to re-draw the outside mold line for the external tank, which in turn is used to make CFD models.

Copyright Publications & Communications, Inc. Jul 2005