Across Acoustics

Student Paper Competition: Modeling Unexploded Ordnances in the Ocean

December 09, 2022 ASA Publications' Office
Across Acoustics
Student Paper Competition: Modeling Unexploded Ordnances in the Ocean
Show Notes Transcript

Over the past 100 years or so, ordnance has entered aquatic environments around the US at military testing and training sites. As these sites transition away from military use, it’s necessary to clean up unexploded ordnance that may have been left behind. We interview Kyle Dalton, one of the winners of the POMA Student Paper Competition from the 182nd meeting of the ASA, about his research regarding the modeling of these unexploded ordnances so they can be detected with sonar.

Associated paper: Dalton, Kyle. Simulating elastic targets for sonar algorithm development. Proc. Mtgs. Acoust. 46, 070002 (2022);


Find out how to enter the Student Paper Competition for the latest meeting.

 Read more from Proceedings of Meetings on Acoustics (POMA).

Learn more about Acoustical Society of America Publications.

Music Credit: Min 2019 by minwbu from Pixabay.

Kat Setzer (KS)


Welcome to Across Acoustics, the official podcast of the Acoustical Society of America’s Publications office. On this podcast, we will highlight research from our four publications, The Journal of the Acoustical Society of America, also known as JASA, JASA Express Letters, Proceedings of Meetings on Acoustics, also known as POMA, and Acoustics Today. I'm your host, Kat Setzer, Editorial Associate for the ASA.


Today we'll be talking to Kyle Dalton about his article, “Simulating elastic targets for sonar algorithm development,” which appeared in the 46th volume of the Proceedings of Meetings on Acoustics and was one of five winners of the POMA Student Paper Competition for the 182nd meeting of the Acoustical Society of America, which took place this past May in Denver, Colorado. This episode is part of a five-episode series highlighting winners of the POMA Student Paper Competition. So Kyle, congratulations, and thanks for chatting with me today. How are you?


Kyle Dalton (KD)


Good. Thank you so much. And thanks for having me on today.




No problem. So first, tell us a bit about yourself. Where are you studying? And what do you research?




Sure. So like Kat said, my name is Kyle Dalton. I'm in my third year at Penn State, working towards a PhD in acoustics. My research focuses on signal processing techniques for synthetic aperture sonar, more specifically imaging techniques to improve the detection of unexploded ordnance. I think like many acousticians, I grew up in music, playing in bands from middle school up through college. I got my undergrad degree in Electrical Engineering and Computer Engineering at the University of Virginia in 2018. And it was while I was at UVA that I think my kind of interest in acoustics started to shift from more of a musical sense to more of a technical sense. My senior design project was a vibration-sensing touchscreen; we also had to write a socio-technical thesis, kind of looking at the interaction of engineering and society. So I wrote mine on how electronic advancements of the 20th century shaped the definition and distribution of traditional folk music, kind of highlighting that shift from music to electronics, and acoustic elements of that. So after graduating, I worked as a computer engineer for the Department of the Navy, installing testing, troubleshooting, and certifying sonar systems. And that was sort of the segue into the underwater acoustics world where I do my research now. 




But yeah, I was gonna say, and here you are today. So your article has to do with target localization using sonar. So can you give us some background on this field?




Sure. So “sonar” is an acronym; it stands for sound, navigation, and ranging. And there are sort of two broad methods for finding a target using sonar. First, you have passive sonar, and that's really just listening for the sounds that an object makes and using that information to figure out what direction it's coming from, and what it is. Think about using thunder to audibly track a storm or picking out someone's voice from a crowd of people. That's passive sonar. And then there's active sonar, and this is probably what most people picture when they think of sonar. You've probably seen this in battleship movies. This is where you send a ping out from your sonar, the sound bounces off of an object, and then an echo comes back to you. Folks might also have experience with some natural forms of active sonar; you can think about bats and dolphins using echolocation. 


So with active sonar, we can measure the time between when the ping from the sonar goes out and when the echo comes back. And then if we know how fast the sound is traveling, we can solve for the distance between the sonar and whatever caused the echo. And that's how we do target localization with active sonar. 


My research uses only active sonar. The work in my POMA supports a SERDP, that's the Strategic Environmental Research and Development Program, project to find unexploded ordnance. Over the past 100 years or so, this ordnance has entered aquatic environments around the US at military testing and training sites. Many of these sites are now transitioning away from military use, so it's obviously very important to make sure that we clean up any unexploded ordnance that may have been left behind. This gets tough because the ordnance can become buried over time. So not only do you have to find a bomb that's underwater, but one that's underwater and potentially buried underground underwater.




Yeah, so that sounds like a very difficult challenge and also, as you mentioned, a very important one. So in your research you specifically use synthetic aperture sonar. What is synthetic aperture sonar?




Yeah, so it's probably best to start with the concept of a real aperture, and I think that's a word that a lot of people have probably heard in the context of a camera or an optical lens. With a camera, you adjust the size of your aperture to control how much light gets in. And you sort of have the same idea in sound, where the aperture is the effective length of your sonar array. Loosely speaking, the longer your aperture is, the better your sonar’s resolution is. But making a sonar array that's really long gets complicated, and expensive. So one solution is to make a small aperture and mount it on a moving object, then you move this array along a path, gathering data as you move. So send out a ping, receive an echo, then move a little bit, receive an echo or send out a ping, receive an echo, move. And as you move, you build up what we call a “synthetic aperture” that's much longer than the physical length of your array, then you can use some kind of clever signal processing to stitch the pings together to form a high-resolution image. 




Okay, yeah. So what are the problems with current reconstruction methods?




Sure. So when I say reconstruction, that just means that we're turning acoustic data from our sonar into pictures. So current synthetic aperture sonar reconstruction methods are really good at making pictures of rigid stationary objects. So they're really good at producing high-resolution imagery of things like the sea floor or shipwrecks. A quick Google search of synthetic aperture sonar shipwreck will show you all sorts of really neat images. The problem with unexploded ordnance, is that they're they aren't perfectly rigid. Under the right conditions, when you ping them with sound, you can excite these structural modes, that causes sort of a sort of ringing behavior. Think about hitting a musical cymbal or a triangle; you'll get some sound back that reflects off of the ordnance, but then you get this sort of this ringing that carries on in time. In my research group, we call this “late-time energy” because the ringing reaches the sonar later in time than the initial returns. And current reconstruction methods don't focus this late-time energy very well. So in the reconstructed image, you get this blurry swath of energy underneath your unexploded ordnance. And there have been some attempts to focus similar types of acoustic features in sort of adjacent related areas. But nothing that addresses this late-time energy with operational constraints and complexities that we were with when looking for unexploded ordnance.




Okay, yeah, that comparison of the cymbal is really interesting, and It's understandable how the blurriness is such a problem. So what were the goals of this study?




Yeah. So this work was really meant to sort of put some modeling tools in my toolbox, if you will. My research is in improving reconstruction techniques, not really so much in target modeling. But I need data containing these elastic objects in order to develop better reconstruction algorithms. Going out on a boat with a real sonar and looking at real targets gets logistically complicated, and you're limited to whatever targets you have available. And, you know, the Navy's not thrilled about having unexploded ordnance out and about for people to test with. So it made sense to make a model to simulate these elastic targets as sort of a foundation that I could build my research on. So I was listening to an episode of this podcast to try to get an idea of what it was like, and one of your previous guests said something to the effect of “a model should be as simple as possible, but no simpler.” And that was really kind of my approach for this, for this work. I needed something that captured, or rather, I didn't need something to capture the exact nuance and every physical phenomena of these elastic targets. I just needed enough to start building a reconstruction technique that I can then refine once I am able to move to real data. Unexploded ordinance is roughly cylindrical. So the model that I chose, which models elastic cylinders, seems like a pretty logical choice.




Yeah, that totally makes sense. So how did you develop your model from there?




Yeah, my code is largely a combination of two separate, previously developed models. The cylinder model comes from a series of papers written in the late 80s by Timothy Stanton, and then the sediment—so I'm modeling what the cylinder looks like on the seafloor, so you need a sediment model—and the sediment is modeled using a point-based sonar scattering model that was developed at Penn State. A point-based model treats the seafloor as a bunch of individual points and then simulates how sound interacts with all the individual points and then combines all the results together. So then there's additional code that I wrote that defines your properties of the sonar, tells the sonar where to drive, how often to ping, how the cylinder is oriented, sort of the like wrapper code, if you will, to bring the two models together.




Okay, so how do you deal with targets that aren't rigid in your model?




The target scattering, especially for non-rigid targets, can get very complicated very quickly. It's a function of the object's shape, size, material composition, the angle you ping it at, the frequencies you ping it at… I could go on and on. And there are folks out there who dedicate their entire career to looking at the scattering of sound off of cylinders. So thankfully, I was able to build on a lot of previous work. Stanton's 1988 paper on the scattering of sound from elastic cylinders integrates the volume flow per unit length of the scattered field for an infinitely long cylinder over a finite length. So in English, he just sort of derives all the math for a theoretically infinitely long cylinder, and then shortens it to a sort of tangible cylinder that we might see in the real world. So thankfully, he derived all the math. So I just had to translate the math to code, and then write some additional code to configure the cylinder size, material composition, and add in the sonar geometry and setup. 




Okay, cool. So then how did you verify your model?




Yeah, I started out just trying to recreate the plots that are in the paper that I was working from. Well, I was able to, you know, take my plot and compare it to the plot in the paper, and build some confidence that what I was doing was on the right track. Then, once that was good to go, I checked the target strength that I was getting out of my model versus the target strength of other models. There are several books out there that give you the target strength of ideal rigid cylinders that made a great comparison to the elastic cylinders that I was simulating, using this new model. Then I made what we call “acoustic color plots,” and that shows how the target strength of the cylinder changes with frequency and with angle, and compared that against, what I was getting, against theoretical acoustic color plots. Then I started making images using my modeled data. Everything kind of passed the eye check, and then I started comparing my images against real images of unexploded ordnance and just kind of subjectively seeing, you know, how did it look? How close is it?




Does it actually look like it's supposed to? 



Yeah, yeah, exactly. 



Um, so what did you find most interesting or exciting in this project?




Oh, man. I think getting that first positive result, like, where it, you know, it looked like a cylinder. It was supposed to be a cylinder. And it's, I think so often in this type of work, you can be 95% of the way there, and it looks absolutely awful. And then you spend, so you spent so long trying to get that last 5%. And so I think when that first result came in, it's like, oh, my gosh, like that is a picture of an elastic cylinder – We did it! I think that was especially exciting.




Yeah, I can imagine that would be very, very satisfying. So what are the next steps in your research?




So I think the big takeaway from the POMA was that my model appears to be capturing the relevant late-time effects of an elastic cylinder on a volume of sediment. And so the next, the next step is using the data from the model to develop a method for focusing the late-time energy that I talked about earlier. And since submitting the POMA on this modeling effort, I think I've come up with a focusing method that works pretty well. So now I'm switching over from modeled data to data that's collected with physical sonar systems. I'll do some analysis to see how well the focusing technique that I developed works on real data and how it quantitatively improves the sonar image. Longer term, I'll be looking at the impact of focusing late-time energy on our automatic target recognition using machine learning. So I expect to have some exciting results for ASA Chicago next spring. 




Yeah, that is exciting! I can't wait to hear about it at the Chicago meeting. 


Well, thank you again for taking the time to speak with me today about your research. It sounds like you've gotten to do some really important work that will hopefully help with finding those unexploded ordinances which you know is probably something many folks even outside of underwater acoustics research can appreciate the value of. And of course, once again, congratulations on winning the award from POMA.


For any students or mentors listening from around the time this episode is airing, we're actually holding another Student Paper Competition for the most recent meeting in Nashville. So, students, if you presented at the national meeting, now's the time to submit your POMA. We're accepting papers from all of the technical areas represented by the ASA. Not only will you get the respect of your peers, you’ll win $300, and perhaps the greatest reward of all, the opportunity to appear on this podcast. And even if you don't win, this is a great opportunity to boost your CV or resume with an editor-reviewed proceedings paper. The deadline is January 8, 2023. We'll include a link to the submission information on the show notes for this episode. 


Thank you for tuning into Across Acoustics. If you would like to hear more interviews from our authors about their research, please click subscribe or find us on your preferred podcast platform.