Inside The J-2X Doghouse: Engine Control — Open versus Closed Loop

As I was driving to work this morning, I came up over a rise and saw suddenly appear in my windshield, over towards the left on the other side of the road, a police cruiser with a radar gun mounted in the window.  Even before I could think about it, the pressure of my right foot on the accelerator lessened.  I then instinctively looked at the speedometer and found that I was traveling 48 miles per hour on a road for which the speed limit is 45 mph.  Thankfully, the officer apparently forgave me the 3 mph violation and continued to wait where he was for a better opportunity to serve and protect the community. 


What I find interesting about that little episode was the immediate, unthinking response I made in response to seeing the cruiser.  In terms of control systems, that could be called a feedback loop.  My senses received data, my brain rapidly processed that data and then sent a signal to react to the data, and then my calf muscles responded by easing up on the throttle.  The car, in turn, slowed to respond to the lower throttle.  Never mind, of course, that if I’d been really, really speeding all this would have been too late (and I would have arrived at work very angry), there was nevertheless a closed-loop response that is not too dissimilar to what we do for rocket engine control … in some cases.


First, let’s get familiar with a couple of terms –


Open-loop: We typically refer to something as open-loop when we have instrumentation that measures conditions in the engine but the engine itself does not respond to those measurements.


Closed-loop: We typically refer to something as closed-loop when we have instrumentation that measures conditions in the engine and then, potentially, the engine takes action based upon those measurements.


It is based upon those definitions that I would call my response to seeing the cruiser closed-loop since I responded and did something with the data.  Another example would be the more modern systems that are used to monitor and control automobile systems.  It used to be that you had a temperature measurement stuck in the coolant loop.  You could watch the temperature rise, but until the system went kaput and boiled over leaving you stranded alongside the road, there was no active, closed-loop control.  Nowadays, if the computer in my pickup truck sees that the engine temperature is too high, it will take action to try and protect itself.  For example, it will inhibit the use of the air conditioning system since that represents an additional power requirement on the engine.

 So, what do we control on a rocket engine?


One thing that we control is power level.  On the Space Shuttle Main Engine (SSME), power level is controlled in a closed-loop manner.  This means that the main combustion chamber pressure is measured as an indication of thrust level and in response to that measurement a valve is opened or closed to increase or decrease the engine power level.  On the J-2X, power level is controlled in an open-loop manner.  This means that we measure the main combustion chamber pressure but we don’t have any feedback loop where we control a valve to ensure that we’re on target.  Instead, should we happen to be off on power level, we have to physically change an orifice in the engine between tests.  The “feedback loop” is data analysis and a guy with a wrench.  Which approach you choose to take are dependent upon your requirements of performance and affordability. 


Another thing that we control on a rocket engine is the mixture ratio (i.e., the ratio of oxidizer to fuel).  Given that on a rocket you are carrying both your oxidizer and fuel with you in the vehicle, you certainly want to make sure that you consume your propellants in the correct ratio to get the most uumph out of them.  Again, on SSME we control mixture ratio in a closed loop manner.  There is actually a small flowmeter on the SSME and, using the data from that flowmeter (and some associated calculations), we move a valve on the engine to dial in the correct mixture ratio.  It’s a pretty nifty system.  Also again, on the J-2X, we have an open-loop system for mixture ratio just like we have for power level.  We test the engine, look at the results, and, if necessary, make a physical change to the engine in the form of an orifice.


Because of these two areas, power level and mixture ratio, SSME is usually referred to as a “closed-loop engine” and J-2X is usually referred to as an “open-loop engine.”  Now, this terminology is not entirely correct since there are some closed feedback loops within the J-2X control system pertaining to engine health and status diagnostics, but we all know how enduring shorthand designations can be.  Also, engines don’t have to be one or the other.  They can be half-and-half.  The engine used on the Delta IV vehicle, the RS-68, sort of falls in this category. 


How you choose to design your engine control system is driven by your requirements.  Put real simply:  The SSME is all fancy-schmancy because it had extremely tight power level and mixture ratio precision requirements and because it was a reusable engine.  The J-2X is intentionally more simplistic because it has looser precision requirements and because it is expendable (and throwing away orifices is a whole lot cheaper than throwing away valves if your requirements will let you get away with it).  Requirements drive design.


Note that I will save the fun topic of engine diagnostics — and the potential for long philosophical meanderings within that realm — for future posting. 




Let’s end this posting with a fun little exercise.  Above is a simplified schematic of a gas-generator cycle engine kind of like a J-2X.  I have shown in the schematic two orifices #1 and #2 (highlighted in yellow).  With those two orifices, we can calibrate the engine.


Scenario:  Power level too low, i.e., measured main combustion chamber pressure too low.

·         Solution:  Increase the size of orifice #1.

·         Explanation:  By increasing the size of orifice #1, I will deliver more oxidizer to the gas generator.  This will deliver more power to both turbines thereby increasing how much propellant gets pumped into the engine.  More propellants pumped in equals more thrust and greater overall power level.


Scenario: Power level too high, i.e., measured main combustion chamber pressure too high.

·         Solution:  Decrease the size of orifice #1.

·         Explanation:  The exact opposite of the previous scenario.


Scenario:  Mixture ratio too low, i.e., the flow of oxidizer is too low in proportion to the flow of fuel, as measured by the test facility.

·         Solution:  Decrease the size of orifice #2 and decrease size of orifice #1.

·         Explanation:  By decreasing the size of orifice #2, I decrease the amount of flow that is diverted around the oxidizer turbopump turbine.  I therefore increase the flow through the turbine thereby increasing pumping power of the oxidizer side.  So I increase oxidizer flow to the engine.  However, by increasing oxidizer flow to the engine and doing nothing else, I’ve probably messed up my overall engine power level so I’ve got to back down a little bit by decreasing the size of orifice #1.


Scenario:  Mixture ratio too high, i.e., the flow of oxidizer is too high in proportion to the flow of fuel, as measured by the test facility.

·         Solution:  Increase the size of orifice #2 and increase size of orifice #1.

·         Explanation:  The opposite of the rationale for the scenario immediately above.


See, being a rocket scientist isn’t that difficult, really.  Now you too can calibrate an open-loop rocket engine.



P.S., I read in the paper this morning that NASCAR racer Kyle Busch had his civilian driver’s license revoked for 45 days for doing 128 mph in a 45 mph zone.  Well, at least I wasn’t going that fast when I came over the rise this morning and saw the police cruiser.  Then again, I wasn’t driving a $400,000 Lexus LFA sports car like Mr. Busch was…





J-2X Doghouse: Okay, So We Ain't That Smart — Yet

Welcome back to the J-2X Doghouse.  We’re going talk about some test results and test data, exactly what Data Dogs love most to do.

Back in the day — back before I had the carefully regulated, federally mandated, and strictly enforced lobotomy that allows people into the ranks of management — I was once an analyst.  And, since it seems so long ago that it doesn’t sound like bragging anymore, I will admit that I was pretty good at it.  I absolutely loved the process of using fundamental physics or empirical correlations for fluid dynamics, thermodynamics, and heat transfer all together to simulate in computer coding how things function in the real world.  Whereas many people enter the field engineering because they like mechanical things or electronic things, there are some of us who relish the seeming purity of problem solving in abstraction.

 Over the years, working on many diverse projects and building many diverse mathematical models to simulate many diverse systems, I came to the realization that my models always appeared most unassailable and brilliant when there was no test data against which to compare them.  To put it bluntly, test data always proves that you simply ain’t as smart as you thought you were.  But, that’s okay.  If that wasn’t the case, then you wouldn’t bother to test.  The whole point of testing to gather data and learn more.

With all due respect to the Serenity Prayer, this ought to be the analyst’s prayer relative to testing:

Grant me —
— the results to validate that which I do understand
— the data to explain that which I did not understand
— and the openness to accept that I can always understand better

That last line is critical.  Ignoring data contrary to what your model output is a seductive, addictive, and dangerous path to follow.  We don’t/won’t do that.

That brings us to the subject of test A2J003 of the J-2X development engine E10001.  This was our first test to mainstage operation.  The planned duration was to be seven seconds.  On Tuesday 26 July, right around five in the afternoon, the test ran for 3.72 seconds and then shutdown.  We did not accomplish the full duration.  Why?  Basically because we ain’t as smart as we thought we were.  We had analytical models telling us that performance would be X, but the hardware knew better and decided on its own to perform to Y.  Here is a cool video of the test:

J-2X Engine Test A2J003

A more detailed explanation of what happened is that the engine shutdown due to the measurement of a pressure too high in the main combustion chamber.  The measurement crossed a pre-set “redline” and the test controller unit took the automatic (and autonomous) action of shutting down the engine in a safe and controlled manner.  The high pressure in the main chamber was caused by higher than expected power coming out of the oxidizer pump.  This, in turn, was due to more power being delivered to the turbine side than expected.  It comes down to a fluid dynamics phenomenon (pressure drops) and what we have is not inherently bad, just different than expected.  So, in essence, we used our models to predict that the pressure in the main chamber would be at a certain level — indicating a certain power level — but the different performance of the hardware resulted in pushing us away from our analytical prediction.

  • Here is the good part:  We learned something.  We learned that our model needs to be updated and we collected the data that will allow that to happen.
  • Here is another good part:  We got enough data, despite the short duration, to recalibrate the engine for the next test thereby making it far more likely that we will hit our target. .
  • Here is yet another good part:  We had a successful demonstration of the test controller redline system by safely shutting down the engine.  The engine looks fine.  The controller did exactly what it was supposed to do and protected the hardware.  In fact, for these early tests we have the redlines clamped down pretty tight specifically to protect the hardware as we learn more about the engine..
  • And here is, finally, yet another good part:  Other than the power applied to the oxidizer turbopump, most of our other predictions with regards to hardware performance appear to be awfully darn good.  So, we’ve got a preliminary validation for much of our understanding of the engine.  Indeed, this is a brand new engine and we have just accomplished mainstage operation in the second hot-fire test.  That is truly unprecedented..
  • Here is the bad part: We have to spend a few minutes explaining to folks not directly involved that despite not achieving full duration, the test was in reality a total success.

If that, then, is the bad part, I can live with it.  I can live with admitting that we ain’t as smart as we thought were.  Why?  Because now, after the test, we are indeed smarter.  And we will continue to get smarter and smarter about the J-2X design until, one day, we will be smart enough to say that, yes, we understand this engine so well that it is safe enough to propel humans into space.