X-33 Composite Tank Problems

On January 20 of this year, Lockheed Martin announced a slip in the X-33 program because of a flaw in one of the two liquid hydrogen tanks. Both fuel tanks are composite, and the defect appeared during an autoclave cycle.

Initial reports on the problem were sketchy, and in some cases just plain wrong. Since then, however, more details have been reported, so it's easier to see the magnitude of the problem.

What Is the X-33?

The X-33 is an experimental launch vehicle being developed by Lockheed Martin under contract to NASA. Unlike currently flying launch vehicles, it only requires a single stage to reach orbit. This concept is known as SSTO. (Other vehicles currently under development are also SSTO.)

The X-33 looks a bit like the Shuttle orbiter, but it doesn't have a large external tank or boosters. It looks and flies a lot like an airplane, so it is often called a space plane. Unlike the Shuttle, the X-33 is fully reusable.

The X-33 is really a technology demonstrator. Lockheed Martin's ultimate goal is to build a larger version of the X-33, called the VentureStar, as a shuttle replacement and for launching large payloads.

More details about both vehicles can be found on Lockheed Martin's VentureStar page.

Fuel Tank Description

The liquid hydrogen tanks are two of the largest structures in the X-33 vehicle. A cutaway view of the internal configuration shows that the tanks occupy well over half of the internal volume.

To conform to the shape of the vehicle, the tanks are not simple cylinders. They are made from four lobes which taper towards the forward end of the vehicle. The lobes and domes are connected by longerons and frames. Each lobe is capped by a dome. As the overview page states, these are not easy structures to build.

Composite Manufacturing Process

All of the tank shell pieces are a sandwich structure, made from graphite/epoxy facesheets and a graphite/epoxy honeycomb core. The inner facesheet has seven plies; the outer has fourteen.

Alliant Techsystems lays up the skins on a fiber placement machine, then precures them. They are then post-bonded to the honeycomb core using a film adhesive. Alliant performed the first bonding procedure, but Hexcel performed all subsequent bonding operations.

Alliant delivers the completed wall segments to Lockheed Martin. The tanks are then assembled in a series of up to eight autoclave cure cycles. Beyond eight cycles, the strength of the laminate can begin to degrade.

During the fifth cure cycle of one of the tanks, on 23 December 1998, a large delamination was found in one of the wall segments. The flaw was sometimes described as "bubbles and cracks" in the "inner lining." This was one of the segments bonded by Hexcel (early press reports stated the walls were made by Alliant).

Failure Analysis

Early reports of the failure lacked a lot of detail and in general seemed to be inaccurate. In its January 25 issue, though, Aviation Week & Space Technology ran an article which had a good amount of detail. The primary source for the article was T. Cleon Lacefield, the Lockheed Martin Skunk Works program manager for X-33.

The analysis which follows is mainly based on details from that article.

Contamination

One possible cause of a delamination is material contamination, and Lockheed Martin is examining that possibility. Contamination of any of the primary materials--the core, the film adhesive, or the composite--could promote a delamination.

Contaminants can be dirt, solvents, moisture, or other liquids (such as oil). Film adhesives and prepregs are processed through strict quality programs, so the introduction of contaminants during material manufacture is unlikely. But both materials are kept frozen until use. If they are thawed incorrectly, moisture can condense and be trapped in the part.

This moisture can show up as bubbles or blisters during subsequent cure cycles, which was reported in some of the early articles. However, such bubbles usually show up after only one or two thermal cycles.

Another possible contaminant is an actual physical object. Release papers are sometimes left on a layer, which prevents any bonding. A razor blade or other small object left in the laminate can be an initiator of a disbond. Such objects, though, are easy to detect, and probably would have shown up during inspection.

A more likely source of contamination is the core. In general, cores aren't as carefully controlled as prepregs or film adhesives prior to use. Also, because of their geometry, they are better dirt attractors (and holders).

Because cores can easily get dirty, compressed air is often used to clean them out. If the air isn't clean and filtered, it may contain oil, which would be deposited on the core. Such contamination would weaken the bond between the film adhesive and the core.

Core Venting

Each cell in a honeycomb sandwich is an airtight vessel. When heated, the air in each expands, increasing the pressure. If the pressure gets too high, the film adhesive bond may fail, initiating a delamination.

The Aviation Week article emphasized this as a possible failure mode, but it is an unlikely one. As stated in the article, the loads created under these conditions should be much lower than the adhesive allowables. Also, the autoclave pressure would tend to counter the effects of the internal pressure (i.e. external pressure also increases). Of course, with a pre-existing flaw (such as contamination), internal pressure could initiate a delamination.

Venting the core would require two modifications: use of a perforated core; and a vent path outside the part. Perforated core has small holes between the cells so air can flow within the core. A vent or vents must also be provided through the skin to provide an air path to the outside.

Lockheed Martin decided not to vent the core because the cryogenic temperatures could cause trapped air to liquify. This alone is not a significant problem, but the resultant drop in pressure could draw in additional air, which would also be liquified.

During flight, the vehicle weight would increase. Once the temperature rose above cryogenic temperatures, the liquid air would return to gaseous forms. Core venting is a relatively slow process, so internal pressures could temporarily become very high--perhaps even higher than in an unvented core during an autoclave cycle.

Ultrasonic Inspection

The tank walls were inspected using an unspecified ultrasonic method prior to the Lockheed Martin autoclave cycles.

Ultrasonic inspection works by transmitting a sound wave through the part. When the wave hits an interface, part of it is bounced back to the surface. By measuring the delay in the return of this signal, the depth of the interface can be determined.

If the interface is a bond (between plies or between the facesheet and the core), most of the signal continues through the part. If there is an actual gap at the interface, however, most of the signal is reflected and the remainder is lost. Such a gap can occur at the back of the part or at a delamination.

Ultrasonic inspection can detect actual gaps or changes in material (such as a foreign object), but it is not good at detecting flaws such as porosity or "kissing" delaminations (a weak or nonexistent bond where the parts actually touch). Furthermore, the minimum flaw size which can be detected is usually about 0.25 inches in diameter.

The latter limitation means that the probe head must inspect the part at intervals no greater than 0.25 inches. Ultrasonic inspection, therefore, is a very slow process. The inspection process can be sped up by determining the maximum size of an acceptable flaw. For example, if a 0.50 inch flaw won't cause the part to fail, then 0.25 inch flaws are of no concern.

Finally, ultrasonic inspection by itself tells nothing about a part's strength: it only tells whether or not a flaw is present. Tests must be conducted on sample panels with known flaws to determine the effect of a flaw.

Precured Skins

One thing not questioned in any of the recent articles is the basic manufacturing method. In particular, the practice of precuring the skins and then bonding them to the core should be examined.

Many honeycomb sandwiches are made by cocuring the skins and the core. One skin is layed up on the tool, the honeycomb is placed against it, then the other skin is layed up on the core. The entire sandwich laminate is then cured in an autoclave cycle. Film adhesive may or may not be used.

This process achieves an excellent bond between the facesheets and the core. Before cure, the facesheets are flexible, and they conform easily to the core. A tight bond is achieved over the entire sandwich.

With precured skins, the core must conform to a rigid surface. If the core can't conform well enough, the gap may be too large for the film adhesive to form a proper bond. Such a bond may not be detected during ultrasonic inspection, and it could lead to delaminations.

Cocuring skins does present some problems, but they are not insurmountable. Handling can be more difficult, but that must be weighed against the difficulty of achieving a uniform bondline.

Cocuring can also lead to dimpling of the skins, from the autoclave pressure pushing the laminate into the cells. Such dimpling is usually not a structural problem, and can be minimized (or eliminated) by adjusting the autoclave temperature rates and pressures.

Conclusion

It will be interesting to see what Lockheed Martin turns up as the cause of the problem, and what will be done to prevent similar problems in the future. This is a challenging part to build. Successfully completing it will advance the state-of-art in composite manufacturing.

If anyone on the X-33 program is reading this and is willing to contribute additional details, or if you have any comments on this article, please contact me at composite.guide@about.com.

Discuss in my forum