Tyler O'Neal, Staff Editor DEVELOPER TOOLS

Simple, fuel-efficient rocket engine could enable cheaper, lighter spacecraft

It takes a lot of fuel to launch something into space. Sending NASA's Space Shuttle into orbit required more than 3.5 million pounds of fuel, which is about 15 times heavier than a blue whale.

But a new type of engine -- called a rotating detonation engine -- promises to make rockets not only more fuel-efficient but also more lightweight and less complicated to construct. There's just one problem: Right now this engine is too unpredictable to be used in an actual rocket.

Researchers at the University of Washington have developed a mathematical model that describes how these engines work. With this information, engineers can, for the first time, develop tests to improve these engines and make them more stable. The team published these findings Jan. 10 in Physical Review E.

"The rotating detonation engine field is still in its infancy. We have tons of data about these engines, but we don't understand what is going on," said lead author James Koch, a UW doctoral student in aeronautics and astronautics. "I tried to recast our results by looking at pattern formations instead of asking an engineering question -- such as how to get the highest performing engine -- and then boom, it turned out that it works." CAPTION The researchers first developed an experimental rotating detonation engine (shown here) where they could control different parameters, such as the size of the gap between the cylinders. The feed lines (right) direct the propellant flow into the engine. On the inside, there is another cylinder concentric to the outside piece. Sensors sticking out of the top of the engine (left) measure pressure along the length of the cylinder. The camera would be on the left-hand side, looking from the back end of the engine. CREDIT James Koch/University of Washington{module INSIDE STORY}

A conventional rocket engine works by burning propellant and then pushing it out of the back of the engine to create thrust.

"A rotating detonation engine takes a different approach to how it combusts propellant," Koch said. "It's made of concentric cylinders. Propellant flows in the gap between the cylinders, and, after ignition, the rapid heat release forms a shock wave, a strong pulse of gas with significantly higher pressure and temperature that is moving faster than the speed of sound.

"This combustion process is literally a detonation -- an explosion -- but behind this initial start-up phase, we see a number of stable combustion pulses form that continues to consume available propellant. This produces high pressure and temperature that drives exhaust out the back of the engine at high speeds, which can generate thrust."

Conventional engines use a lot of machinery to direct and control the combustion reaction so that it generates the work needed to propel the engine. But in a rotating detonation engine, the shock wave naturally does everything without needing additional help from engine parts.

"The combustion-driven shocks naturally compress the flow as they travel around the combustion chamber," Koch said. "The downside of that is that these detonations have a mind of their own. Once you detonate something, it just goes. It's so violent."

To try to be able to describe how these engines work, the researchers first developed an experimental rotating detonation engine where they could control different parameters, such as the size of the gap between the cylinders. Then they recorded the combustion processes with a high-speed camera. Each experiment took only 0.5 seconds to complete, but the researchers recorded these experiments at 240,000 frames per second so they could see what was happening in slow motion.

From there, the researchers developed a mathematical model to mimic what they saw in the videos. 

{media id=236,layout=solo}

"This is the only model in the literature currently capable of describing the diverse and complex dynamics of these rotating detonation engines that we observe in experiments," said co-author J. Nathan Kutz, a UW professor of applied mathematics.

The model allowed the researchers to determine for the first time whether an engine of this type would be stable or unstable. It also allowed them to assess how well a specific engine was performing.

"This new approach is different from conventional wisdom in the field, and its broad applications and new insights were a complete surprise to me," said co-author Carl Knowlen, a UW research associate professor in aeronautics and astronautics.

Right now the model is not quite ready for engineers to use.

"My goal here was solely to reproduce the behavior of the pulses we saw -- to make sure that the model output is similar to our experimental results," Koch said. "I have identified the dominant physics and how they interplay. Now I can take what I've done here and make it quantitative. From there we can talk about how to make a better engine."

New space weather advisories serve aviation

Tyler O'Neal, Staff Editor DEVELOPER TOOLS

A new international advisory system is working to keep aircraft crew and passengers safe from space weather impacts, thanks in part to the efforts of a team of CIRES and NOAA developers, forecasters, and scientists in Boulder, Colorado.

"Thunderstorms or snow can disrupt flights. What we call 'space weather' can be disruptive, too," said Rob Steenburgh, a scientist and forecaster in NOAA's Space Weather Prediction Center (SWPC). So when the United Nations' International Civil Aviation Organization solicited interest from space weather centers around the globe to create a new advisory system for aviation, SWPC offered to help, on behalf of the United States.

"Space weather" generally refers to the changing conditions in space and in the Earth's atmosphere that result from activity originating on the Sun. Solar "winds" can ebb and flow, buffeting Earth's magnetic field, and when a coronal mass ejection (CME) passes by Earth, it can seriously shake up that magnetic field. Such "space weather" can temporarily disrupt navigation and communications systems and can even increase radiation levels in the atmosphere. CAPTION NOAA's Space Weather Prediction Center in Boulder, Colo. CREDIT Photo: Katie Palubicki/ CIRES & NOAA National Centers for Environmental Information.{module INSIDE STORY}

And--much as with a blizzard or looming thunderstorm--airlines can take action, changing an aircraft's flight path to avoid or lessen impacts, said CIRES solar physicist Hazel Bain, who worked on the project for the last year. A moderate radiation advisory, for example, might motivate airlines to divert planes from the polar regions, she said, since flights over the poles are more exposed to radiation than flights at lower latitudes. CIRES is a partnership of NOAA and the University of Colorado Boulder.

In late 2018, ICAO chose SWPC to be one of three forecast centers from around the world to begin issuing space weather advisories to the civil aviation community. Just 12 months later, SWPC's system was operational.

"This new capability will permit flight crew and flight operations experts to make use of the most updated information possible on any solar events which could potentially impact aircraft systems or passenger health," ICAO Secretary General Dr. Fang Liu said in a statement late last year.

The project focused on three main types of impact on aviation: Disruptions to satellite-based navigation; disruptions to HF radio communications, and radiation levels (see box below).

About 15 CIRES and NOAA scientists, software developers and forecasters spent the last year racing to design a system that would meet ICAO's specs--that organization decided, for example, what levels of radiation would constitute a "moderate" event and what "severe" disruption to global navigation satellite systems means.

For each type of impact, the team needed to develop or integrate models of the Sun-Earth environment, to support aviation 24/7. For radiation, for example, they brought in a Federal Aviation Administration model, Bain said. For satellite navigation warnings, they quickly finished a model that had been in the works already (ROTI or Rate of Total-Electron-Content Index). And for high-frequency communications, they leveraged another existing model, the CTIPe, for Coupled Thermosphere Ionosphere Plasmasphere Electrodynamics.

That was tricky enough. But the project also demanded careful attention to the user interfaces and e-tools forecasters would need to help them issue advisories on current conditions, and forecast conditions 6, 12, 18 and 24 hours out.

"We built intuitive interfaces by working directly with the forecasters," said Ben Rowells, a CIRES developer in SWPC. Forecasters are interested in a space weather phenomenon called ionospheric scintillation, for example, which can impact navigation. "A forecaster can select hotspot regions on an animated data map and rapidly issue an advisory," Rowells said.

Since late November, the SWPC team has been taking two-week turns with the other teams--both international consortia that developed independent advisory systems--to issue advisories.

"This was revolutionary," said NOAA's Steenburgh. "While there is still more work to be done, the models and user interface were made available to SWPC forecasters through a heroic effort from virtually all parts of SWPC."

In the three months since advisories began, no space weather center has issued "moderate" or "severe" warnings in any category, Steenburgh said. This lull won't last, however, as we move towards the next solar cycle maximum, estimated to arrive in about four years.

Army researchers develop efficient distributed deep learning

Tyler O'Neal, Staff Editor DEVELOPER TOOLS

A new algorithm is enabling deep learning that is more collaborative and communication-efficient than traditional methods.

Army researchers developed algorithms that facilitate distributed, decentralized and collaborative learning capabilities among devices, avoiding the need to pool all data at a central server for learning.

"There has been an exponential growth in the amount of data collected and stored locally on individual smart devices," said Dr. Jemin George, an Army scientist at the U.S. Army Combat Capabilities Development Command's Army Research Laboratory. "Numerous research efforts, as well as businesses, have focused on applying machine learning to extract value from such massive data to provide data-driven insights, decisions, and predictions."

However, none of these efforts address any of the issues associated with applying machine learning to a contested, congested and constrained battlespace, George said. These battlespace constraints become more apparent when the devices are using deep learning algorithms for decision-making due to the heavy computational costs in terms of learning time and processing power. A networked set of agents (denoted as colored nodes) train their individual deep neural nets using locally available data while interacting with neighbor nodes through available communication links (represented using grey edges).{module INSIDE STORY}

"This research tries to address some of the challenges of applying machine learning, or deep learning, in military environments," said Dr. Prudhvi Gurram, a scientist who contributed to this research. "Early indications and warnings of threats enhance situational awareness and contribute to how the Army evolves and adapts to defeat adversarial threats."

The researchers presented their findings at the 34th Association for the Advancement of Artificial Intelligence Conference on Artificial Intelligence in New York. A pre-print version of the paper is online.

In an earlier study (see Related Links below), the researchers demonstrated that the distributed deep learning algorithms can yield the same performance as the typical centralized learning algorithms without aggregating the data at a single, central location while decreasing the learning time linearly with the number of devices or agents involved in distributed learning.

"Distributed learning algorithms typically require numerous rounds of communication among the agents or devices involved in the learning process to share their current model with the rest of the network," George said. "This presents several communication challenges."

The Army researchers developed a new technique to significantly decrease the communication overhead, by up to 70% in certain scenarios, without sacrificing the learning rate or performance accuracy.

The researchers developed a triggering mechanism, which allowed the individual agents to communicate their model with their neighbors only if it has significantly changed since it was last transmitted. Though this significantly decreases the communication interaction among the agents, it does not affect the overall learning rate or the performance accuracy of the final learned model, George said.

Army researchers are investigating how this research can be applied to the Internet of Battlefield Things, incorporating quantized and compressed communication schemes to the current algorithm to further reduce the communication overhead.

The Army's modernization priorities include next-generation supercomputer networks, which enable the Army to deliver leader-approved technology capabilities to warfighters at the best possible return on investment for the Army.

Future efforts will evaluate the algorithm behavior on larger, military-relevant datasets using the computing resources available through the U.S. Army AI Innovation Institute, with the algorithm expected to transition to run on edge devices, George said.