Math helps NASA’s spacecraft modification

Math helps NASA’s spacecraft modification 

A mathematician from the Worcester Polytechnic Institute (WPI) is working to make NASA spacecraft lighter and more damage tolerant. For this, he combines 19th-century mathematics with cutting-edge machine learning. Carbon nanomaterials form spacecraft structures like rocket fuel tanks. Mathematics can give a great hand to detect any imperfections.  

Randy Paffenroth is an associate professor of mathematical sciences, computer science, and data science. In this research project, he has shown his multi-dimensional ability.  He found that an old mathematical equation is very helpful when he combined it with machine learning and neural networks. The combination was resulted in finding an algorithm. This increases the ability to scan the density more accurately. The resolution level gets higher and higher while using this technique. For instance, the higher resolution scans are nine times "super-resolution" 

that can detect imperfections in Miralon® materials ( a product of the Nanocomp Technologies, Inc. and they are flexible, powerful, lightweight nanomaterials).  

Another spacecraft nano-structure is Miralon® yarns that are as tiny as human hair. It is used to wrap the spacecraft structures like fuel tanks so that they can tolerate heavy pressure. This is a very crucial aspect of the space journey and any flaws or thickness imperfection can be a serious issue. Paffenroth has made a team with graduate students. They are analyzing data about how the products are being manufactured. Then they focus on to make them more consistent. 

To scan the nanomaterials, Nanocomp uses a "basis weight" scanning system. It is a modified and commercial process to detect mass uniformity and imperfections. It provides a visual image of density at the end. The research team tries to train their algorithms by using machine learning. It will increase the scan process’ capacity to get higher resolution images. The algorithm is a unique one. The "compressed sensing / super-resolution" algorithm increases the resolution by nine times.

The algorithm was built with the Python programming language. The artificial neural network is the basement of this algorithm. Paffenroth has "trained" the algorithm on thousands of sets of nanomaterial images. He found many imperfections in the results. He also used this in a series of practice tests to get the "ground truth." He says "I give it a sheet of material. I know the imperfections going in but the algorithm doesn't. If it finds those imperfections, I can trust its accuracy,"

His algorithm was combined with the Fourier Transform to make it more effective to scan a high-resolution image out of a low-resolution image. Fourier Transform is one of the earliest mathematical tools that breaks down an image into its individual components.

"We take this fancy, cutting-edge neural network and add in 250-year-old mathematics and that helps the neural network work better," said Paffenroth. "The Fourier Transform makes creating a high-resolution image a much easier problem by breaking down the data that makes up the image. Think of the Fourier Transform as a set of eyeglasses for the neural network. It makes blurry things clear to the algorithm. We're taking computer vision and virtually putting glasses on it. It's exciting to use this combination of modern machine learning and classic math for this kind of work."

Nanocomp Technologies funded Paffenroth's work. The company is a major player that makes advanced carbon-nanotube materials for aerospace, defense, and the automotive industry.

Spacecrafts are already using Miralon®. NASA's Juno probe is a great example of where it has been used recently. It orbits the planet Jupiter. The new carbon composite pressure vessels is a prototype of the next-generation rocket fuel tanks. NASA has used Miralon® on them too. It increases the spacecraft’s strength and durability.   

At present, Nanocomp is working on enhancing Miralon® yarns three times stronger as a part of NASA’s current contract. If this comes to light, Paffenroth's team should be appreciated for their effort in this project. 

The Quality Manager at Nanocomp, Bob Casoni says, "Randy is helping us achieve this goal of tripling our strength by improving the tools in our toolbox so that we can make stronger, better, next-generation materials to be used in space applications. If NASA needs to build a new rocket system strong enough to get to Mars and back, it has a big set of challenges to face. Better materials are needed to allow NASA to design rockets that can go farther, faster and survive longer."

The responsibility of Paffenroth’s team now is to develop algorithms in a way that it can predict the workability of a piece for the first time. It will be done by using active feedback control systems ensuring to build a more consistent end product. The algorithm analyzes the manufacturing from the beginning to the end. 

Paffenforth says, "We can use machine learning to predict that Nanocomp won't get a useful length of material out of a particular production run. It helps them with waste. If they can tell in the first few meters of the run that there will be a problem, they can stop and start over. The Holy Grail of process engineering is that the more you understand about your process, the better your process is."

The results of the research were published by WPI on Aug. 25 in 2019 in the International Conference on Image, Video Processing, and Artificial Intelligence in Shanghai, China.