When we consider the advancements of data acquisition technology, drones, or unmanned aerial vehicles (UAVs) have become increasingly common for data collection.
I've heard dozens of examples of how UAVs have been used to identify damage on elevated components, such as flare tips. I've seen probes mounted onto these platforms to provide ultrasonic spot checks at high elevation, safely. I've seen many sites collect lidar data, mapping out extraordinarily accurate positional information on equipment across a large site.
Huge, interactive maps or even visual models of oil platforms-- this is some of my favorite imagery that's been created in the inspection industry lately, and it's something I'd like to explore further.
Near Earth Autonomy is a Pittsburgh company adding a whole new level of functionality to these aerial systems. Alex Foessel and Near Earth are building systems that allow for the increased deployment of these technologies with automation.
Below are excerpts from my conversation with Alex. Listen to the full podcast!
QUINN
Tell us about what brought you to Near Earth.
ALEX
I've known the founders, most of them for many years, more than 20 years. Once I decided that I wanted to land in Pittsburgh, we agreed that I would evolve the technology into a product portfolio that would make the business successful from a product standpoint.
QUINN
And what is the product?
ALEX
Near Earth has two main areas of technology.
One has to do with mobility, which is all about the autonomy in the aerial economy for transporting cargo, for example, critical elements such as medicines, or even people. And that has both a commercial variant as well as one that has military implications.
Where I'm putting all of my time is inspection.
One of the key things that I would like to highlight is that inspection is a business of today. Many people and a lot of resources are being spent every day to maintain, inspect and guarantee the up-time of very expensive assets.
Connecting with my history at John Deere, managing customer support was all about delivering uptime. That is the most important thing and that is really what makes or breaks the profitability of potential customers.
So we’re trying to create a business around inspection, aerial inspection.
QUINN
You're absolutely right. The equipment that we're inspecting represents major capital investment.
What are the core technologies that you're bringing to the field that sets you apart?
ALEX
Fundamentally, Near Earth develops aerial autonomy. But when you apply aerial autonomy to inspection, it’s the ability to fly in a precise, repeatable, and safe way, using the same inspection path over and over and over, as it relates to an asset.
For example, the assets could be tunnels, oil tanks, construction sites or aircraft. So there's a number of potential assets that require consistent repeatable inspections.
QUINN
That's fantastic-- reproducibility or repeatability of the inspection and being able to track it over time is key.
ALEX
If you do a spot check on an asset, certainly you will know something.
Many times, inspections are just spot checks, but you may want the ability to repeatedly and precisely capture a coherent data set.
For example, you want the images to be taken from the same point of view so then later you can say, last year, the surface looked like this, so then you start establishing trends and thinking of predictive maintenance. Then you can understand, “hey, this crack is growing or it’s still the same,” or “we need to worry.”
Then you can at least start making a more targeted maintenance plan and do the maintenance at the time when downtime is not a big issue, because it's planned downtime, as opposed to having to shut down the asset, creating all kinds of costs.
QUINN
You're giving the clients the tools to get ahead of the curve and to really anticipate that stuff. And that's fantastic.
ALEX
We can connect our autonomy inspection system to different types of types of drones. Each application may be different.
We also have different types of sensors. In some applications, our technology using lidar to localize is also the sensor that maps to establish certain, let's say, geometric characteristics of your asset. In other cases, it's a camera that takes very high resolution photos that could be used for visual inspection, without exposing a human to risk.
QUINN
I was going to ask about that and tie the drones to very specific applications.
ALEX
The first application is the inspection of existing assets to find out whether there’s deterioration. The second is the inspection of our ongoing construction sites. And the third one also has to do with inventory. If you have an inventory in a large warehouse, you want to fly the same path over and over to be able to assess, does your warehouse management system contain the truth as to where you have all the boxes?
So these are three types of inspections-- inspections of assets, inspections of processes and inspections of inventory.
QUINN
Those are some very special applications and to my understanding, the drone autonomy currently is tied very closely to GPS, so how do you maneuver GPS-denied navigation?
ALEX
Algorithms take the information from spinning lidars and construct a new map while localizing the drone with respect to that map.
Once you have the first flight, then you have the map, and from then on, you can repeatedly fly the same path using this point cloud that is generated using the lidar.
QUINN
Oh that's fascinating. Is there a range of how accurate it is?
ALEX
It's between two to three centimeters, so that is the level of accuracy you can get. So whenever you fly again, and you're taking photos, you know that photo is being taken from the same location, within a couple of centimeters.
QUINN
Perfect. I had to ask because I see it as crucial to reproducibility.
You've mentioned a number of fascinating applications. Are there any other capabilities that may have surprised you?
ALEX
There's a couple more that are not as mature. Once you have a drone that can precisely localize itself and fly close to a surface, we're also looking at the possibility of deploying contact sensors. In those cases, for example, you could be testing for the integrity of carbon fiber, or you can test coatings.
This is preliminary, but I see that capability being an important part of our roadmap.
QUINN
Being a refinery-focused guy, I could really see flare stacks being able to collect some UT data in addition to some visual data on a flare stack tip, for example.
But it's really taking it to the next level, though, with the level of reproducibility.
ALEX
The other aspect I believe is relevant here, is when you need to deploy a group of inspectors to do a manual inspection, in addition to the safety risks, in addition to the complexity of deploying people, you're only going to do those inspections every so often, when they're needed.
I'm not saying that things are under-inspected, but you’re certainly not going to over inspect them, considering that complexity.
Once you have a system that can actually fly and deploy and follow this path, this is opening the door to executing much more frequent inspections that may change some of the inspection paradigms.
You may be able to consider more frequent inspections. And now we also have another dimension, which is how we're starting to build data.
QUINN
Machine visual learning, do you see that as a component to this?
ALEX
Yes. There's always this trade off between what is really cool and exciting and what is practical, right?
If you can establish that your drone-enabled inspections are equivalent to that of a manual inspection and the inspector sees that he is safer, more comfortable and more effective by looking at images on a screen, then I say that's the first and practical step.
Then we can talk about machine learning and all those things. I don't want to downplay them, but to me, it’s like, let's get the basics in place. Then we can talk about next steps and building more refined algorithms and tools.
QUINN
Getting the images in front of the inspector, in comfortable working conditions. That's where it all starts. That's the immediate application and I have no doubt about that.
ALEX
One thing that we could talk a little bit about is the distinction between autonomous versus manually flown drones. Because the manually flown drones already deliver a good element of value by taking the inspector out of harm's way.
But it’s more difficult for a manually flown drone to deliver this repeatability for spot checks. I'm not being critical, this is one of the steps of creating value for clients. Once you start building this database of inspection data over time, days, months, years and whatnot.
You can actually make sense of all that data. Then autonomy is going to deliver that precision.
QUINN
When we look at data output, are there any integrations with existing reliability and maintenance systems?
ALEX
Not only is it important to capture the data and build a database for each asset, but also to integrate this data into their existing inspection and maintenance systems.
QUINN
With new data management systems, processes like data acquisition are going to be feeding into larger and larger data sets. It’s important to integrate those key findings.
ALEX
At the end of the day, inspection data has no value unless it's actionable. So the actionability is key, right?
When you look into what types of decisions you're actually making with that data, that may also trigger feedback to the team that does the inspection, so you have this continuous improvement of the inspection, to drive decisions.
