About SDHQ | Learning to Dive | Nitrox | Wreck | Deep Cave | Deco | Eanx | HOTx | Trimix
PADI | NAUI | IANTD | Schedule | Articles | Reviews | Charts/Tables | Gallery |
Dive Centers | Destinations | Organizations | Information | Shop Online
The Truth About Cracks and Scuba Tanks
by Matthew MacLean
The following is very long and attempts to answer as many question as possible about Al scuba tanks, cracks and testing mechanisms. It is a follow-on from the " It was a Aluminium tank that blew" thread (in Usenet Newsgroup "rec.scuba") by Skip Commagere of E Force fame, after a series of posts on rec.scuba discussing aluminum tank failures and existing techniques to detect such failures. It refers to several links and other posts from that thread, but does not require reference to them for understanding.
It is written as simply as I can make it but it is still a fairly technical subject. To that extent I apologise to people who have a good understanding of this field for the simplifications inherent.
What is 6XXX aluminium?
6XXX grade Al is a wrought alloy aluminium. Wrought alloys differ from cast alloys in that they can be shaped by deformation.
It is an age hardenable ternary alloy. All "ternary" means is that it has 3 primary constituents, these being Al, Si and Mg. It may also include trace elements of both Cu and Cr. 
"Age hardenable" means that the material can be treated to produce a fine, hard, coherent precipitate in the softer more ductile matrix. More on this later.
A typical 6XXX grade alloy is 6061-T6. This is the current industry standard grade for aluminium tanks.
Composition of 6061- T6
1% Mg, 0.6% Si, the rest Al with trace elements of Cu and Cr. Typically these trace elements are less than 0.5% combined. 
So what does the T6 stand for?
T6 is the designation for the type of material treatment process that the Al has undergone. T6 stands for solution treated artificially aged. 
Solution Treated Artificially Aged
To achieve this, the material is raised to a temperature above room temp to allow different phases within the material to form. At some point when the desired precipitates form (i.e. come out of solution,) the alloy is quenched to lock the new microstructure in place.
Over Ageing occurs when a material is held at a process temperature past its optimum material properties. At this point after quenching the yield and tensile strength are lower than expected. This may also occur at room temp due to diffusion but the time periods are huge and depend on the precipitates formed. Over Ageing will also occur if a material temp is elevated at any point in the material’s life.
6XXX series alloys are usually raised to 190 degrees C and held at that temperature until the desired material properties are achieved and the metal is then quenched in hot water (80 degrees C.) 
Typical properties of 6061-T6 aluminum include a tensile strength of 45000psi and a yield strength of 40000psi. This is in vast contrast to pure aluminum (annealed), with a tensile strength of 6500 psi and a yield strength of 2500 psi .
The material is basically a matrix of Al with Mg2Si precipitates spread evenly throughout it. It gains its strength from the interaction of the "hard" Mg2Si precipitates with the soft Al matrix as dislocation attempts to take place when the material is under stress.
Why is aluminum used in scuba tanks?
The primary design characteristics for a pressure vessel material are a high yield strength and high fracture toughness (K). More accurately, we wish to maximise the material ratio properties M1 and M2 as given below. 
Fracture toughness (K) is a measure of the materials resistance to cracking and is a property of the material the same as yield strength or hardness. Typical valves of K for Al/Mg/Si alloys (T6 treated) are 25 Mpa/m2. 
Secondary material considerations which may or may not have validity when designing scuba tanks are final weight, corrosion resistance and cost of fabrication.
Materials that exhibit the primary material selection behaviour are some Al alloys, steels, Cu alloys, Ni alloys, some Ti alloys and other specials, and GFRP (carbon fibre and like). 
After we consider manufacturing costs and corrosion issues as part of the design consideration, we are left with the choice of steel or aluminum for our tanks.
In practice, weight consideration for aluminum and steel tanks is not as great as you would think, but the buoyancy characteristics alter significantly for the two classes of tanks, and this comes into consideration for many users.
So, where is the problem?
When metals are subjected to cyclic loading at stresses far less than the material yield strength, they can fail in a catastrophic manner. This is called fatigue . (The point of using fracture toughness of the material in its selection from a design point of view is to limit this event.) The difference between aluminum and steel tanks is that steels have an "endurance limit". Stresses applied to the steel below this limit will never create fatigue. Aluminum, on the other hand, has no endurance limit, so even small repetitive stresses will eventual cause failure as the metal fatigues.
Does this mean that aluminum tanks crack and steel tanks don't?
No, due to the level of stress induced in service, both tanks’ materials will develop cracks eventually, its just that the time periods are different. We don't see cracking as a major issue in steel tanks because other failure modes occur faster (i.e. rusting) and tend to remove the tank from service before cracking becomes a problem. Also, hydrostatic tests are more effective at picking up cracks in steel due to the slower crack growth rates (more about this later), and the better chance for plastic deformation of the remaining uncracked material during a hydro test.
Scuba tanks are designed to one of two design philosophies, these being "leak before burst" and "yield before burst". These philosophies have to do with the cracking nature of the material used and the stresses under which it operates. Basically they say that any crack in the tank should either result in the tank deforming before it fails catastrophically (i.e. explodes) or it should leak before a crack will get to the size that will result in catastrophic failure (i.e. the crack length required for catastrophic failure is longer than the tank’s wall thickness.) There will be a lot more on this later.
Why is this of interest to scuba divers?
Its important, because certain older aluminum alloy tanks have not done any of the things that the designers have intended (i.e. leak or deform) and instead have blown up, severely injuring people. This is called catastrophic failure.
To understand this we need to understand fracture mechanics and crack growth theory.
Crack Growth and Fracture Mechanics (as it applies to scuba tanks)
Cracks form when localised stresses in the material exceed the plastic deformation limit of the material. Cracks can initiate from many things, including foreign inclusions, surface defects, micro voids, rust, manufacturing induced stress concentrators (sudden changes in wall thickness, etc.) Whether one of the above mentioned factors results in a crack depends on local stresses, orientation and a bunch of other things. The same mechanism in one spot will start a crack and in other act as a crack inhibitor.  All we really need to know here is that the conditions for crack formation will be meet in service at some point and then the crack exists.  At this point, cracks can either be non-propagating or propagating. In the course of this discussion I will concentrate on propagating cracks.
Generally once in existence, the crack will grow relative to the energy applied each cycle, the geometric properties of the crack itself, and certain material characteristics like the size, number and distance between mirco voids in the material. The more voids the faster the crack will grow for a given energy/crack geometry configuration.  Cracks in this state exist in everything from bridges and ships to nuclear reactors,  and scuba tanks.
So what happens next?
Cracks in this state continue to grow until one of three things happens:
1. ... the remaining uncracked material yields due to the increase in stress resulting from effectively less material carrying loads. (Your tank fails hydrostatic test due to plastic deformation). This is a result of the "yield before burst" design philosophy, coupled with a proof test.
2. ... cracks continue to grow until they exit the surface of the tank and hopefully you notice a leak (pin hole leaks in tanks are often noticed during fills in baths). Or, cracks can be detected visually/remotely during hydrostatic and visual inspection. This is the "leak before burst" design philosophy.
3. ... the crack continues to grow until the critical strain energy release rate condition is met, and the tank blows up.
Relative to this discussion, it is event 3 which interests us.
(A quick note on the two design philosophies mentioned above. Only one of them is used by designers, and they apply it to cracks running in one direction (usually straight through the wall), but occasionally due to geometry a crack will run around the inside of the tank, and then the other can come into play through luck more than good management, since critical length of surface cracks can be quite long.)
In order for the tank to blow up, it must meet the following requirement: 
Kic: is the material fracture toughness described earlier (critical point.)
Pi (p) is the mathematical constant 3.1415…
Stress (s) is the stress present in the material due to its current loading.
"a" is the dimension of the crack in the principle direction (i.e. direction its growing)
Once this equation is met, the crack suddenly grows through the material at speeds approaching the speed of sound, with the resultant release of energy.
From this equation you can see that the guiding requirements are stress and crack size. More stress means a smaller crack size will meet this condition.
The equation is the same for both aluminum and steel tanks, just the material properties are different
This is not quite the whole story, as there is a shape factor that goes into the equation. It is dependant on crack starting point/orientation/type, but becomes a constant for the crack in question. Typical shape factors can be found in .
Also engineering design factors for stress concentrator factors can be introduced here for design purposes (a typical one is a nipple in a pressure vessel raises the local stress (induced) given by a hoop stress equation by a factor of 3 in the area of the nipple fitting). 
Tanks are designed so that this condition can't be met; either the wall thickness is such that the crack goes through it and leaks or the tank deforms first (depending on design philosophy). Currently, due to hydrostatic tests, visual inspections and other means, tanks that are approaching these conditions are detected and removed from service. This is what is supposed to happen; tanks do not have an infinite life. So why did those tanks explode?
6351- T6 alloy, sustained load cracking (SLC)
and corrosion cracking (CC)
6351-T6 alloy displays a nasty property called "slow crack growth." Both SLC and CC are subsets of slow crack growth theory . What happens here with SLC is that once the local K has been exceeded and our normal crack propagation starts, the value for K in the local area drops to less than the design engineer thinks and material testing indicates. The best analogy I can come up with is static friction and dynamic friction. In your car you stop quicker when you hit the brakes if the wheels keep turning cause you are braking using the static value of friction between the tires and the road. Once you lock up the wheels you are using the dynamic valve of friction and the car takes longer to stop because it is lower than the static valve. Something like this happens with SLC; it results in effective crack length a lot longer/quicker than we would otherwise expect. But it resets itself every time you empty the tank. So the next time you fill the tank you have the higher valve for K present, until the crack starts. The tank still has to meet the higher valve of K for it to explode (I think: I'm not 100% sure on this bit; it has to do with the energy model for crack growth/plastic deformation in the local region around the crack tip and total crack energy.)
So why did the tank explode in the E-force shop?
My guess is that the tank did have a fairly standard sustained load crack that was not detected. The reason for the initial confusion put out by the DOT inspector about a new type of “fast fracture” is that it did not seem that the energy requirements had been met for a fast fracture as described earlier. I propose that the increase in stress in the tank came from thermal conditions imposed when the tank was placed in the filling chamber. From what I understand E-force used a cooled water bath to fill tanks in and this created enough thermal stress in the tank to tip it over the balance of the strain equation. This increase in stress can be quite high, especially if the tank had been in the sun for a period of time and suddenly introduced to water at 30 to 35F. (13 to 16 degrees C). I am not implying that thermal shock resulted in the explosion but that the energy requirements for Fast Fracture were met due to a combination of thermal stress and internal stress due to the tank being almost full at the time of being placed in the bath.
I am not in any way implying that procedures at the shop caused this accident. I am just trying to explain the mechanism, this was really just the straw that broke the camel’s back.
What I haven't mentioned so far is why neck cracks are so bad. In the side walls of a tank a design before leak philosophy results in leaks, but in the neck due to the way tanks are made the material thickness can exceed the thickness that a design before leak concept would indicate. Coupled with possibility of incredibly high stress raisers and events that occur during filling, such as tank abuse, etc., the potential for the fast fracture condition to be met exists as a statistical probability.
Corrosion cracking [5, 3] is similar to SLC but uses the higher valve of K (the one we think the material has), and comes about because environmental elements and residual stresses combine to continue crack growth after normal propagation methods have stopped. In the case of aluminum tanks, one agent is Cl- ions found in seawater. Really, this just means to wash your tanks, and inspect them for pitting (I got this bit off some dodgy web page. J )
Dag Debeitz's 5283 grade Al tank
In one way, this one is even worse. The material properties of 5283 aluminum are believed to alter in relatively short time periods. Since K is a material property and gets lower as the material ages, then the length of a critical crack for a “fast fracture” becomes less and eventually will be less than the wall thickness. On the day in question, it was estimated the temperature in the shed was 40 degrees C, so the pressure in the cylinder would have been greater than when filled. Coupled with the lower K value, a crack (that would have been stable when the tank was made) became unstable and blew up. This tank material looks of poor quality even from photos on the web page.
Testing For Cracks
Eddy current test:
If you want to know how they work go to the following site: http://members.aol.com/flare439/myhomepage/visualeddy/Visual_Eddy_Brochure.PDF
What does it mean? Eddy current tests determine changes in resistance inherent in the metal. It is considered a technique requiring high operator skill for correct interpretation.  Due to these consideration it is usually used for go-no-go checks on like components . If one result reads different from all the rest, it is rejected. An eddy current tests is good for picking up surface and subsurface defects. Due to the fact that it detects changes in electrical resistance in the material it will detect cracks, inclusions, changes in material density and get different results for paint and unpainted parts. Because of this, in industry when used as a board-based search technique, it is always backed up by a second technique, usually ultrasonic testing, to correctly identify the flaw type .
Having said all this, the manner in which eddy current testing is used for testing for cracks in a tank’s thread area is an almost ideal usage model for the technique. It is repetitive, very localised and tailored for the operation at hand. But it still can't tell you if what you are looking at is a crack or another kind of fault (it can tell you where and depth). If a tank failed testing due to this, after several years of testing by the same technique, I would have no problem in scraping the tank. On a new tank I would require the tank to be tested by other means to determine whether it was a manufacturing induced crack or other defect from manufacture or just a statistical anomaly in material thickness in the tank.
Note, this technique would not have detected the crack in Dag's tank, and may not have detected the one in Force-E explosion (assuming it was what was mentioned by the inspector on the web site, an eddy current test would not detected the crack, since the inspector claimed the crack didn't exist prior to the onset of fast fracture mechanism.)
Where does this leave us?
If I owned a 6531- T6 grade tank, I would be getting it hydrostatically tested and then crack tested yearly. Why? Because the hydro acts as a proof test and tells me I don't have cracks of a certain size present as can be calculated from the equation for fast fracture (reverse the equation and solve for "a", put in the hoop stress for the hydro pressure and K from a material hand book and you have critical crack length that is required for the tank to fail.) If the tank doesn't fail, then that length crack doesn't exist. I would then follow up with an eddy test for crack growth in the neck and, since the proof stress is greater than normal operation stress, the tank should be OK for service for its normal cycle life for one year. I would also store it at 40bar. If this is uneconomical in the States, I would be trading it in and buying a new tank.
If I owned a 6061-T6 tank, I would be getting any tanks over 10 years old eddy tested for neck cracks yearly. Other than that, I would ensure hydro's are in order, and make sure the local visual inspections were being done properly. This should give a high level of confidence to users as there exist no reports of 6061-T6 tanks exploding. That alloy does not suffer from SLC.
Note: In Australia tanks are hydro'd yearly so the proof test concept is how this cracks are discovered.
If I owned steel tanks, I'd fear rust; tanks used with high 02 gas mixes can rust at accelerated rates and should regularly be inspected internally, especially if saltwater contamination is a possibility.
Ice divers, if I were you I would look at the ductile-to-brittle transition of BCC (body centred cubic) materials and its effect on K values for steel. (I saw this lovely photo once of scuba tank stored on the ice at about –30 degrees C.) This doesn't affect aluminum tanks.
As for cracks, because steel has a higher K valve than aluminum, cracks in steel tanks with the same stress characteristics and shape factors as in aluminum tanks grow at a slower rate. Hence you have more time to detect them. Remember, the steels will crack eventually if nothing else gets them first. Look at the history of LP versus HP steel tanks for conformation of this.
Why am I happy to say I would condemn a tank
on the basis of an eddy current test?
(if you had followed the originals thread I was critical of this).
Because in the world of equations and testing I can determine the length of the crack I have, the length of the crack needed to go critical, and the number of cycles required to grow the crack to that length and the amount of useful life left in the tank.
Unfortunately in the neck region some very complex actions take place and the one thing I cannot be certain of is the amount of actual stress being applied to cracks in that area. So in this situation, once I have detected a crack (and made sure it is a crack,) I will happily scrap the tank, because the real stresses in this region can be far higher than predicted. (Ross Bagley put up a nice model showing some of the complexities involved.)
The next time someone scraps one of your tanks, be happy; he just made your day!
Just for clarity, I have no intention of giving up using aluminum tanks, but I might look at them a little more closely in future.
Matthew MacLean Mackay Qld
A few loose points from the original thread,
1. All metals work-harden to some degree, just the amount varies.
2. Comments that steel “barrels before bursting” is a result of design issues only and not a product of material property as implied by the original poster. Once the energy requirements are met, the material will fast fracture no matter what it is.
About the Author
I am a practising Mechanical Engineer, with a degree (B.E Mech.) from the University of Newcastle NSW, Australia. I have my own consulting business (MacLean Engineering & Management Services) and I primarily work in the area of Maintenance Management and the development of maintenance systems for capital-intensive industries. I studied materials, scheduling systems and maintenance management in my final years at university (basically cause I had no idea what I wanted to do) and ended up with a job where I occasionally use all three. I am also am avid diver who has worked as a deck hand and local site guide on a dive boat on the Great Barrier Reef, Queensland, Australia (in my days of from my normal job.)
Thanks to Rich Lesperance for picking up some very ordinary mistakes and typing in the original.
 The Science and Engineering of Materials, Donald R Askeland, 1984 PWS Publishers
 Engineering Material and Their Properties 2nd edition, Flinn/Trojan Houghton Mufflin Company / Boston 1975
 Mechanical Metallurgy, George E Dieter, McGraw Hill Book Co, 1988
 Materials Selection in Mechanical Design, M.F Ashby, Pergamon Press 1992
 Elastic & Plastic Fracture, Metals, Polymers, Ceramics, Composites, Biological Materials. A.G Atkins & Y-W. Mai, Ellis Harwood Publishers 1985. This is very much a definitive text for general use.
 Engineering Materials 1, An Introduction to Their Properties and Applications, Michael F. Ashby and David R.H Jones Pergamon Press 1980
 Maintenance Management, APESMA Short Course Notes, Prof. John Chambers and D. Wilkinson, University Of Newcastle Press, Adapted from the course notes for the Maintenance Engineering Post Graduate Course . 1990
For more information contact us via email