Dr. Vadim Smelyanskiy, Principal Scientist, Ames Research Center, Moffett Field, CA
- Thursday, 27 March 2014
Dr. Vadim Smelyanskiy, Principal Scientist, Ames research Center, Moffett Field, CA
Dr. Vadim Smelyanskiy is a principal scientist for physics-based methods in the Exploration Technology Directorate at NASA Ames Research Center. During his tenure at NASA, he has been the principal investigator on several projects funded by NASA and other government agencies.
NASA Tech Briefs: What project or projects are you working on now?
Dr. Vadim Smelyanskiy: As a principal scientist in physics-based methods, I oversee a number of activities that require this approach. One of them is related to critical physics-based analysis of failure mode development and fault management, as well as assessment of various risks in the integrated vehicle environment.
Basically, when you have abnormal behavior in a flying rocket, the root causes of that and the consequences and the timing, as well as the instrumentation that is needed to predict and diagnose and isolate the causes, are all based on physics. With traditional physics disciplines, somebody is doing composite materials or structural mechanics, and somebody is doing electromechanics, or computational fluid dynamics. In a situation of a complex integrated physics environment, however, where mechanical, propulsion, electromagnetic instrumentations are all tied together, there's no such possible software code that would integrate all this multiphysical simulations in one coherent fashion. There are too many interfaces. It's a huge inhomogeneity of physics processes — a variety of special skills, and diversity of those processes.
It's like trying to describe a human being through one physics model that couples the mechanics of your motion and what happens in the brain down to the neuron level. That's simply not possible. When you want to capture the underlying physical phenomena, it's a critical physics analysis that steps in.
On one hand, critical physics analysis always answers the questions that look like a puzzle. If something happens: Why? You really need to develop an intuition and combine things together. If, for example, you have an explosion, you want to know whether the explosion is a catastrophic or simpler fire. [The space shuttle] Challenger is an unfortunate example.
A number of ingredients that contribute to the strengths and prognosis of an explosion are abnormal, and they happen on a very small scale. If a fire occurs in the engine, and part of it gets into the liquid fuel tank, those things have to be understood through physics. There are a variety of projects that support space-line systems, that support in-space autonomy through the cryogenic propellant loading, and other applications.
NTB: You mentioned abnormal behavior. What can be detected with this physics-based approach?
Dr. Smelyanskiy: Abnormal behavior is something that happens in an engineering system, either due to the malfunction of engineering components or due to the abnormal external conditions. One of the examples of this is when you send a fluid to load a fuel tank through some pressurization procedure, and a spontaneous formation of vapor blockage may form in excessive fashion somewhere in the pipe. When this happens, essentially you're losing your liquid. Liquid is transferred into the vapor. So you have a vapor blockage. What does that mean?
That means that somewhere inside of the pipe – and the pipe could be long, tens of yards, --- somewhere in the middle of that pipe, is a location that is not known to you. It's not immediately obviously with a limited number of sensors how to determine that location. You have a portion of liquid completely converted to a vapor. Now it's quite clear that if you keep pressurizing that pipe, vapor will compress. That means you're losing controllability. Now when the vapor is compressed, you will have a complex phenomenon happen. At the very least, you are not filling your tank. At worst, you could have explosive nucleation, where large portions of the liquid begin are converting to a vapor with a huge rise in pressure.
This is just one example, a pretty mundane situation that will quickly grow into an unstable, explosive situation. Another example, which I already mentioned, is an explosion. For example, people know about Challenger, and that the catastrophe occurred because of a case breach. That occurred in the lower part of the solid rocket motor, and that breach punctured the liquid engine tank. Most of the people assume that because it's a hot gas, it punctured a liquid engine tank. In fact, it was not. A careful physics analysis of the situation shows that, in fact, the explosion occurred some 30 meters above the point where the solid rocket tank was mounted to the rocket, to the core stage of the rocket.
So what happens? Because the solid rocket lost stability, the mount point disintegrated, and there was a rupture of the hydrogen and oxygen tanks. Both liquids – cold liquids –were very far away from the case breach that caused lots of mechanical instability in the first place. Those two cold liquids essentially collided with each other and collided with the surface of the solid rocket motor. That caused the fire. In fact, the collision of the two streams of liquid boiling oxygen and hydrogen with the surface of a solid rocket motor caused the fire. That was a cold combustion. Why does that event cause a fire as opposed, for example, to a more powerful deflagration or even more powerful detonation? That is a question that really has to do with microscopic analysis of what happens with sub-millimeter-size bubbles that collapse and create a huge pressure that give rise to ignition that on one hand can cause detonation or cause a fire. Those things, which are very delicate, is where physics-based analysis plays a huge role.
NTB: What kinds of technology are you working with, to perform this type of analysis?
Dr. Smelyanskiy: The end product of what we do is an analysis itself. It's the predictions and recommendations and modeling and conclusions based on those models: what will work and what will not work, in engineering vehicles. We've been doing this for space-line systems, also before with the Constellation [space program].
We've also developed a physics-based approach to in-space autonomy, to where the technology constitutes essentially online running models with model-based diagnostic capabilities. It's smart software that understands the physics of the system at hand and interprets the signals coming from that system in a way that allows, with a very small number of observations in fairly complex systems, robust predictions. Examples are cryogenic propellant loading, where we're working with a number of teams.
There is another fairly nascent activity that started about a year ago that has to do with a very different technology, a quantum computer, which harvests the power of quantum mechanics to enable computation to solving NASA computational problems. The power cannot even be compared in any usual metrics with classical computation.
NASA is involved with experimenting in quantum computers [that are more reminiscent] of the analog machines and analog proposals for computation in ‘50s and ‘60s, except the analog being a physical device that implements a process with inputs and outputs. The input into that process is a formulation in some form, or an optimization problem in our case.
This quantum engine is composed of quantum elements, called spins or quantum bits, that are coupled to each other in a concerted fashion, displaying some distinctly quantum behavior. The quantum engine operates in a homeostatic fashion. It maintains itself, flows to its lowest energy state, its quantum-mechanical lowest energy state, and then it morphs from the system whose lowest energy state is generic to another system whose lowest energy state reflects the solution of the classical computational problem at hand. This computation, while it only currently involves 512 of qubits, has a huge potential. We already see very small problems that are nevertheless fairly difficult for classical computing, that in fact could be solved by quantum computer.
It addresses many practical NASA applications, from planning and scheduling. We will try to apply the technology to help with autonomous planning, scheduling, and navigation of the rovers on the moon. We are trying to apply it to different technologies with scheduling and navigation.
This also has to do potentially with many other applications, such as fault management. If you have a complex network, and that complex network has some faults, there are a number of possibilities where the same sensor data could potentially lead. You need to pin down a unique possibility, the most consistent with existing sensor data, and that is a very difficult computational task. We are applying quantum computing to that.
One of the future tasks we plan to apply a quantum computer to is analysis of the National Aerospace System, where you actually have to have optimal control of a few tens of thousands airplanes that in real-time has to air traffic control, accommodate for delays, and create a system response in real time, a very hard optimization problem. There are many other examples, including robotics. Intelligent analysis of the data, reconstruction of certain features and properties, like certain shapes of lakes or craters, for example, from complex imagery would be a very interesting task for a quantum computer.
NTB: What are your biggest technical challenges in getting the 512-qubit machine up and running?
Dr. Smelyanskiy: The biggest technical challenge is that this machine does not a have a very good programmability property. To program that system is basically an art. You need to embed one graph of your problem you want to solve into another graph representing the connectivity of the hardware. Those embeddings are not automatic at this point. Some of them can be efficient; some of them are not. Non-efficient embeddings would include a lot of additional qubits to embed a fairly small problem. So, of course, given that the size of the overall quantum machine is limited right now, 512 qubits, that may severely limit what practical applications you can actually solve. So we may need to wait until the machine becomes 2000, 8000, qubits, before that can really address a broad variety of applications.
Our second challenge is to do hardcore physics research to actually understand the power of that machine. One might say: Why is this a NASA problem? Why isn't this an academia problem? However, the point is that the limitations of performance of the quantum machine given by connectivity, also given by the effect of the noise that destroys the quantum coherence of the operation of the machine, are different, and they work differently for different types of applications. This is where we should do research and understand those limitations and perhaps suggestions of how to change the architecture in the future.
NTB: What are the most exciting possibilities with quantum computing?
Dr. Smelyanskiy: There are two possibilities. One of them: The quantum computer will be able to solve some of the hard NASA problems that will make a difference for NASA operation. Of course, that's a challenge, because the problem is relatively small in size yet relatively difficult in solution. On the other hand, it also has to be very relevant to NASA. Finding this interplay is very scientific challenge. That's a huge possibility, if we can explore it and achieve that goal.
The second possibility, in my opinion, which is not related to quantum, but is in my view, very breathtaking: the ability to inject physics-based analysis into the smart autonomous, typically in-space system operations, where we can really do what military call fly-by-wire. Butterflies cannot fly until they explore the motion of the flow of the air. Complex military airplanes, especially with the stealth capability, do not have a very good hybrid fluid dynamics property that allows easy control. So you need to do electronic autonomous control, responding to the winds.
I would like to have that fly-by-wire capability with in-space operations, be it cryogenic propellant loading, solar electric or others. Having this would require breaking a NASA mentality. I believe we need to inject way more of this type of approach that would benefit us in autonomy in space.