The purpose of these remarks is to explain why we are still concerned with issues that emerged around engine and fuel system at the occasion of some incidents during 2015 and 2016. We are not convinced that final closure has been achieved. Our objective was not to write a scientific paper, but to approach the issues in a way understandable for a somewhat scientifically interested non-expert, without becoming too superficial.
Neither Pratt & Whitney, nor LM mentioned Coriolis forces in connection with the 2015 motor incident. I started to follow that lead through the following reasoning:
After the fire of June 2015 and the subsequent grounding of the airplane, the following explanation was given:
A seal in the compressor part of the motor was damaged and the rotor made contact with the housing. Shrapnel penetrated the fuel tanks, causing a fire.
This could happen because the rotor was subject to a larger than anticipated deformation.
The latter was only possible in case of a faulty estimate of the forces working on the rotor. Please keep in mind: this is a hypothesis. But the latter is based on the assumption that Pratt engineers do not make mistakes while doing the static calculations on a turbine rotor. This looks like a rather safe assumption, but it is of course no certainty.
So let us assume an error in the estimate of the forces involved. What forces are applied on the rotor?
An axial force due to propulsion and compression. (always)
Centrifugal forces. (in curves/ direction changes)
Coriolis forces (in curves/ direction changes)
We need to keep in mind that all these forces work simultaneously.
The former three are so simple that mistakes are highly unlikely. But… highly unlikely does not mean impossible.
Coriolis forces are a bit more complicated. This can especially lead to problems with an engine like the F135 that flirts with the limits. The rotor has a substantial mass (I estimate not much less than a ton) with a considerable diameter (about one meter) revolving at a high rate (15.000 rpm). Note: there are a number of different rpm numbers. What appears to be one axle on the sketch below is in reality multiple shafts that turn concentrically within each other at different rpm’s, connected by gearboxes. This has very little impact on the topic under discussion. I only mention it to preempt nitpicking.
Jet engines are, in spite of the apparent simplicity of the mechanical concept, complex objects that partly elude conventional analytical calculation methods. That is why jet engines used to explode quite frequently on their test benches and then still way too often in midair.
Modern CAE (computer aided engineering) and CAD (computer aided design) methods offered a massive improvement here. Nevertheless, especially while sitting in total awe in front of the fabulous pictures on our computer screens, we always have to keep in mind that this comes at a price. We leave the guaranteed certainty of analytical treatment and we enter the realm of modeling. The problem is that we can only perform static computation on relatively simple, geometrically defined shapes. For more complex cases we discovered somewhat over half a century ago a very elegant detour: the finite elements method. Almost certainly, the inventors have been inspired by the so called “Cross” method, in those days already intensively used in civil engineering. What we do is divide an irregular shape (that we cannot calculate) into a high number of regular ones (that we can compute). We then have to describe the boundary conditions and the interactions between all those elements precisely in a mathematical way. Next we have to bring the whole system to equilibrium. The latter is only possible through an iterative process, requiring massive computation. We were thus depending on computers becoming fast enough to achieve a break through. For all practical means and purposes, this happened with Digital Equipment’s VAX in the seventies.
It may sound strange, but that same finite elements method also finds application in a number of very different problems. Some examples: polymer extruders, blow molding, weather forecast and climate models.
Don’t let that scare you away. The uncertainties with jet engines are clearly a lot smaller than in the latter two examples. They are especially very much smaller than previously, before we had these techniques available. But they remain. A couple more things come on top of that.
In such an engine, we use dozens of different materials, and that over a rather extended temperature range. All material properties do not just vary among materials, but also for a given material at different temperatures. Also statically important properties like elastic module and tensile strength do this. Further all these materials have their own termal expansion. Friction coefficients too vary with temperature. There is always, has to be, friction in this kind of machinery. All that needs to be considered correctly in the model.
Last but not least, we deal in these engines with weight in a truly extremely stingy way: no gram too much! That does not only increase the amount of deformation at a given load above usual levels, it also reduces the margin of error to almost zero. We are approaching the limits of the possible rather closely here. We do of course attempt to not overstep any limit. That is unfortunately never easy and not always very clear and straightforward.
The incident was caused by friction between stator and rotor in the compressor part. But there needs to be friction in order to keep the compressor stages separated. So, there was under the given circumstances, only too much friction. Thus there must have been a larger deformation than provided by the design. If we exclude design errors the reason has to be, that larger or other forces than predicted were at work. That feeds the Coriolis suspicion… Stator and housing are not subject to the same Coriolis forces as the rotor. They experience other and much weaker ones.
The subject bladed wheel was designed to withstand the 500ºC it is normally exposed to. Because of somewhat more friction, the temperature rose to 1000ºC and that proved fatal. This lead to fissures, fracture and fragments that penetrated the fuel system. Note: this happened in the so called cold part of the engine, at the third stage of the compressor. Due to the compression the temperature increases markedly, even if one starts with air at -60ºC.
The drawing below shows, enclosed by a red dashed line, the spot where it happened. The rotor is colored grey, the stator yellow.
On the left we see a detail of the seal with the so called ‘knives’ that is very common in turbo machinenery.
How exactly the parts were deformed and where the friction occurred is not completely clear to me. For that I would have to speculate, and I prefer to not do that. It is also unnecessary.
The engineers, although knowing infinitely much more about the circumstances than I do, took exactly the same path I would have taken. They unearthed a very popular idea from last century: the ‘break in’. By running the engine for a number of hours in a controlled regime, they grind a groove at the critical location. This will of course decrease the amount of friction.
But… they did not lift the ban on more than 5g turns that they had imposed after the incident. I would not have lifted it either, because…
That ‘break in’ is surely the best quick fix available under the given circumstances, but it does not represent a final solution.
Worse: The core lesson we need to learn from this incident is: our CAE model of the motor has – for whatever reason – serious flaws. Reality deviates from the predicted, calculated situation! Quite obviously this is life threatening. It is rather trivial that, in this situation, a serious look at the Coriolis forces seems mandated, although this is by no means the certain cause of the problem. There could be dozens of other things going on and creating mischief. I would like to see that unequivocally clarified before flying with the thing at all. That was, of course a political no-no. Hence the 5g restriction. That will certainly decrease the risk to an – arguably – acceptable level while they can keep up the appearance that there is no serious problem. Especially the test program can continue. I can muster a lot of understanding for the latter aspect. However, as long as we do not achieve final clarification of these issues, we are exposed to all kinds of surprises, none of them pleasant.
One cannot exclude that a final and yet relatively unobtrusive solution may be found, although it will certainly not be simple. But it is extremely unlikely that such a solution can be implemented in a very short term. Wise people wait – if they have that option – until that point is reached (or not) to sign contracts (or not). We should however show appreciation for the fact that the situation the Dutch are in is not that simple. We have no reason to brag about our relatively better position, because it was only the grace of our empty coffers that saved us from fast and stupid commitments. Let’s keep it like that!
If today, the software is mentioned as the reason for the still applicable 5g restriction, that is somewhat unbelievable. I cannot imagine a normal decent engineer giving such an explanation. This probably came from marketing.
We received the following information:
When making high g turns, an excess pressure occurs in the fuel tanks that are then subject to deformations. As a remedy there was talk about safety valves. Also in this case one of the measures was a temporary restriction of turns to 5g.
A couple of months later there was another news flash – apparently independent from the previous one – that detached coating (other sources mentioned “crumbling insulation”) threatened to clog the fuel-hydraulic system. This incident was later attributed to a faulty delivery by one supplier. If true this would cast a strange light on the quality effort of the project.
The Fuel System
Let’s have a more thorough look at this. Below you see the schematic that locates the fuel tanks.
The F35 can internally carry 8390 kg of fuel, stored in seven tanks distributed over the airplane.
I know nothing about the shape of these tanks, but one thing is clear: it will not be neat, easy to compute cylinders. Because in such an air plane we deal as greedily with cubic centimeters as we do with grams. Fuel is stored where there happens to be room. This will always lead to more adventurous shapes for the tanks. Because we deal equally stingy with grams, we can expect the minimum possible material thickness to be used.
There are many different ways to construct fuel tanks for airplanes. One way is to fill the tank with polymer foam featuring a so called ‘open pore’ structure. That way, we avoid the occurrence of a voluminous gas phase that encompasses explosion hazards. I do not expect to see this here: it gives away too many cubic centimeters. I guess however that we will have self-sealing tanks. This means that there is a – rather thick – coating present that will, if the inner coating layer is penetrated expand in contact with the fuel and thereby close the leak automatically.
I guess that we will find different kinds of fuel tank constructions, and that at least in the wings and the stabilizers structural parts of the airframe will double as tank wall. The expression ‘carry-through’ seems to indicate that these spaces, with a relatively restricted volume, are not in the first place seen as storage. Here we pursue another goal. Stealth encompasses, compulsively, a relatively unfavorable aerodynamic shape. The latter causes certain spots of the airframe to get rather hot through friction. That leads to a potentially dangerous IR signature, hence the necessity for cooling. For that cooling we use fuel. Thus it seems likely that fuel from the engine feed tank (F3) is pumped to the engine through wings and stabilizers. The content of other tanks would in this case always go to F3.
I suspect that it is not that simple, for no better reason than that I would not design it this simple. The colleagues working on this have an education and mindset that are not too different from mine. In think that in the above story, the liquid velocities are way too small to achieve sufficient heat transfer rates. Also the self-sealing coating, establishing a rather effective insulation layer would be massively in the way. I would thus install a separate system for the cooling of the most critical spots. In that system we need not apply any other coating than what is necessary as protection against corrosion. In case of leaks, we can shut down this system by closing valves. Or, if possible, no coating at all, if we can find a sufficiently light and heat conducting material that is not corroded by the fuel. These are just speculations. They have a certain degree of foundation but alternatives are thinkable.
Let’s have a look at the fuel. I do not know what the F35 consumes. Let’s call it JPx (x is some unknown number characterizing the fuel). Many modern fighters fly on JP8. The basis is certainly the kerosene fraction of crude oil, but…
It is possible that the composition has been altered in order to manipulate the average molecular weight.
There are certainly a number of additives present. Some of these additives are necessary to neutralize the harmful effects of other additives.
Thus: an exploded pharmacy. Why all this ‘hocus pocus’? Several reasons:
In kerosene we find substances that have a tendency to polymerize. This leads to rubber like precipitations clogging filters, spray nozzles or even pipes. Thus we add substances that inhibit this polymerization. There are however other and even more important reasons for manipulations and additives.
These combustions (plural) are extremely complex processes. What really happens can only be realistically represented in a multidimensional space. There are complex problems of reaction kinetics combined with energy and mass transfer. We have to pull all registers of Reaction Engineering to understand what happens here. I feel completely incapable to explain this problem in a halfway acceptable fashion if I cannot use mathematics.
But also without a complete insight it is possible to understand that in such complex problems yield and/or thrust can always be improved with a couple percent. That is achieved with these additives. But… some additives are corrosive for the used materials. That is why corrosion inhibitors (yet other additives) are added.
At any rate, it is not a good idea to bring JPx in touch with, for instance, aluminum without further ado. It also is unnecessary: we have extensive experience with container coatings. Just look at the inside of preserve and soda cans. This is real state of the art and we have abundant know how. If we want self-sealing systems that does of course not suffice: in that case we need a multi layered coating that will be considerably thicker. The adhesion problems are the same.
Now we have to bother with safety valves for a moment.
On the left side, we see one in its most primitive form. If the pressure in the space around ‘Flow’ gets too high, the force on the ‘wing valve’ will compress the spring and product can escape. The force on the spring can be regulated with the ‘Pressure screw’. This determines at what pressure the valve will open.
Primitive is good, because we install safety valves to anticipate situations that we have not foreseen. They are the suspenders on top of the belt, our last defence. Accordingly they have to work under all circumstances. If properly maintained, they will do exactly that. I never saw safeties fail during my forty years long career.
You are never going to see this type in an airplane: too heavy. But the principle is always the same.
What can go wrong with Fuel Tanks
Fuel tanks are not designed to withstand serious pressure: we save weight wherever we can. Accordingly they need to be carefully protected against overpressure and there are thus always safety valves present. There is a lot that can go wrong:
During fueling we can compress gas bubbles, so that an elevated pressure can occur. We do have shuttle lines to evacuate the gas, but it would not be prudent to assume that everything works perfectly as planned all the time. Further I do not know how this works exactly during airborne fueling.
There is always the possibility (by error) to over-fill the tank in which case we apply the maximum pump pressure to the system. The tank will definitely not be able to withstand that.
These are only the most common cases. Nothing of that kind happened so far. I assume that all precautions against these possibilities have been taken.
Let’s look at what the fuel tanks do during the flight.
In the most primitive case, we are looking at a cylinder and we fly straight at constant height.
In the gas phase we have the normal pressure that was there after fueling.
At the bottom of the tank we have that same normal pressure plus the minor (because of the limited height) hydrostatic pressure caused by the liquid.
Let’s now fly a right curve at 9g.
In the gas phase we still have the same pressure. The gas phase is unproblematic.
In the liquid phase the pressure will increase to:
Normal pressure plus the hydrostatic pressure of the liquid now subject to a much higher acceleration. Thus the pressure has increased.
Still not dramatic. Due to the restricted distances in the airplane, the pressure will be hardly higher than 1 bar. But that the tank has to withstand, of course.
As we want to save weight we will probably keep the tank wall thin and fix it to the airframe structure at multiple points. The attachments, all together, now must (at 8390kg fuel load) withstand 75 ton! Still no big deal.
But what if, instead of a neat cylinder we have a somewhat more erratic shape. And that will be the case in the real world.
Here an example with a protuberance and a gas bubble in it.
What if we fly the same 9g right curve now?
Statically there is not much going on.
We just have that gas bubble G that is now compressed to the same pressure as the adjacent liquid.
But dynamically, there is a world of difference.
The liquid has moved! (red arrow). If the maneuver happened suddenly (which, I believe will normally be the case) the liquid has been ‘pushed’ with a force of 75 ton. Even on its short journey it will collect quite a bit of kinetic energy, which it can only unload on the poor little gas bubble. There will thus, intermediately, occur a pressure that can be many times the normal equilibrium pressure for G. Further, there is a serious shock rocking the fixations of the tank. The shock wave can also potentially deform the tank wall.
Afterwards, everything returns of course nicely to equilibrium… if nothing has been broken in the mean time.
This is the dreaded ‘water hammering’, also called ‘water shock’. This phenomenon has already frequently, when people were opening valves a little bit too enthusiastically, destroyed brand new piping systems with something as innocent as water, to the immense amazement of the ignorant perpetrators.
I suspect this is what happened. And if that is the case, dear colleagues, that was an absolute beginners error. How lucky that there was no tank rupture during the flight!
What to do about this?
We can determine all the points in the fuel system where, during random maneuvers, such trapped gas pockets can be formed. We can then place safety valves in all those spots.
This will of course restrict the overpressure in the gas phase but not the liquid impulse that is still going to come, although seriously attenuated. This is a solution, but by far not a perfect one. But apparently, this is what they are doing.
A perfect solution could only consist of a complete redesign of the fuel system. In that case it is possible to completely exclude de emergence of trapped pockets. However, this leads to a drastic decrease in fuel capacity and/or a complete redesign of the interior space of the airplane. None of both are attractive options that could be realized in a short term.
The fuel system will, at least for a considerable time span, remain a cause for concern. Plastic deformations threaten structural integrity. But also elastic deformations come with the risk of fatigue failure.
And yet another detail: Deformations interfere with the adhesion of all kinds of coating to the tank walls. The latter can get detached, crumble and start to create havoc in filters and spray nozzles. Apparently something of that kind already happened. This will further hoover over our heads a a potential nuisance.
It is of course entirely possible that I am mistaken about all this. But I do not believe that I put an unfair burden on the shoulders of those that advocate the F35, as it is so terribly simple to prove me wrong. Fly unrestricted 9g turns: that is all it takes to silence me!
Gerard De Beuckelaer