# How a Gulf Coast Refinery is Rethinking Inspection and Incident Response with Docked Drones - Event: NestGen ’26 - Published: October 2, 2026 - Length: 29:14 - Watch: https://nestgen.org/library/watch/4FT_TWBrM4U/ > The transcript of a recorded NestGen session, for reading or for use with an AI tool. It comes from the video's > captions; automatic captions can mishear names and terms. ## About this session Tank farms present a constant need for inspection coverage, rapid verification, and safer response when anomalies occur. Autonomous drones are giving operators a way to improve visibility across these environments while reducing the need for manual checks in time-sensitive situations. In this NestGen session, Bronto will share how autonomous drones support tank farm inspection and rapid event response at a Gulf Coast refinery, and what oil and gas teams should think through when deploying similar models across storage and terminal operations. ## Transcript **[0:00]** This session is going to be on how a Gulf Coast refinery is rethinking inspection and incident response with docked drones. And with me to present this, we have Wade Lychee. He's the digital site lead for Bronto Industrial. Wade is the person who actually flies, tests, builds the program on ground. Like he's the chief pilot of that organization. Security patrols or like a fence inspection or a tank inspection or rooftop inspections after, you know, heavy rain or a hurricane. He's the one who is leading everything. What I really appreciate about Wade is that he's honest about where the program really is. So today isn't just like, you know, it's a victory it isn't a victory lap. It's about how this refinery you know, that's already highly automated finds gaps that drone can and we can, you know, build a case for scale within that refinery. So **[1:05]** yeah, also if you find this session valuable, then please tag us on LinkedIn, share some posts, write about us, but yeah. That's about it. And thank you so much. And Wade, could you please share your screen? And yeah. You can start. >> Sure, thank you. Hello everyone. Thank you for your time. As Deep stated, my name is Wade Lesh. I'm a digital site lead for an oil refinery along the Gulf Coast of the US. I help manage the site's drone program that includes both docked drones and non non non docked drones. I also provide support for other digital efforts. Today I'd like to discuss how autonomous drone technology fits into the modern refinery environment. Most refineries have invested heavily into automation over the past several decades. They have sophisticated control systems, extensive instrumentation, advanced historians, and a growing network of sensors. **[2:06]** Yet, despite all those investments, there remains a significant gap between what we measure digitally and what we physically observe. Uh today we'll walk through where those gaps exist, the use cases we identify for the dog drones, and how we discover them, and where we believe the technology is going forward. Uh I'll start off by giving a quick overview tour of a typical downstream oil refinery. Uh a typical oil refinery begins at the tank farm, where crude oil is received and stored before being routed to crude and vacuum distillation units. These units use large fractionation columns, some of the tallest structures in the refinery, uh to separate the crude into different streams based on boiling point. The the resulting feedstocks are then processed through conversion units such as FCCs, hydrocrackers, reformers, uh cokers, which break down and upgrade the hydrocarbons into valuable fuels like gasoline and diesel. Uh there's also hydrotreaters and sulfur recovery units, which remove contaminants and sulfur to meet environmental specifications. **[3:08]** And throughout the refinery, there's extensive pipe racks that serve as the transportation network for carrying hydrocarbons, hydrogen, steam, water, and utilities between process units. There's also supporting infrastructure such as cooling towers, boilers, power generation systems, flare stacks, and water treatment facilities that provide the energy, cooling, safety, and environmental [clears throat] controls needed to keep operations running. Uh the final products are blended and stored in tanks and shipped to customers, completing the transformation of crude oil into usable fuels and products. So, that was just a quick overview tour of a of a typical downstream refinery. Um and as you can imagine, these processes require a lot of heat energy and pressure and a lot of moving parts that that require oversight. So next let's take a look at how the oversight is done today. Uh when people outside of the industry think about refineries, they often imagine large industrial facilities with lots of moving parts that require lots of manual **[4:09]** work. Uh the reality is very different. Today's down downstream refinery is already one of the most automated industrial environments in the world. Uh control rooms monitor thousands of process variables in real time. SCADA systems track equipment status continuously. Uh historians capture years of operational data and sensor networks report pressures, temperatures, flows, and levels every every second of the day. So operationally, refineries are already very good at measuring process conditions. Uh the question is whether they're equally good at observing physical conditions. And that's where the gaps begin. Uh let's talk about how visibility is covered. Today in many instances, vis- visibility in a refinery of assets is achieved only through manual inspection. Uh for example, our site security team runs several vehicle patrols a day that must cover about 5 mi of perimeter fencing. And in some instances, the fence line is not clearly visible from the patrol path to clearly detect if there are breaches or issues with the fence line. Uh so there's certainly a gap where both with both **[5:12]** efficiency and coverage. And second is achieving visibility at elevated heights of structures in a refinery. Uh in many operational settings, this is only achieved today by sending an operator or an inspector out into the field to physically climb these structures for visibility where they must take precautions to protect themselves from hazardous issues and physically climb one tank at a time. Uh so so these visibility these visibility gaps can lead to under- detected issues. Uh a tank level transmitter may tell us that a tank is functioning normally, but it doesn't tell us whether water or product is accumulated on a roof maybe after a rainy event or or some such event. A structural sensor doesn't tell us if corrosion is developing in a specific location. The control room cannot visibly inspect a flare tip. Most refineries have hundreds of tanks, miles of fencing, and countless elevated structures that remain dependent on periodic human inspection. >> [clears throat] >> These assets are certainly managed, but they're not continuously observed. So, while we often describe **[6:15]** refineries as highly automated, there's still a gap between automated measurement and continuous visual awareness. This is the gap we're starting to address with the Doc Drones. So, we've developed so far we've developed three primary use cases for the Doc Drone missions. First was inspection, second was security, and third you see listed here as emergency response. Inspection tends to be the most obvious use case. It usually has the It usually has the highest financial impact from drone observation just depending upon what is detected and what impact that has on costly approaches. Emergency response is usually well understood. Security is often least discussed, but it may ultimately provide some of the most compelling value, but we'll walk through each of these cases individually. First, let's look at inspection applications. Inspection is where we most of the operations team initially saw the clearest value. Many of our critical assets are elevated, remote, or difficult to access. Tank roofs are a good example of that. Inspecting them often requires operations to travel **[7:16]** significant distances and physically climb each structure for for visibility on top of it. And as you can imagine, a few operators can take significant amount of time to climb and visually check dozens of tanks while also increasing their safety risk at working at heights and being introduced to hazardous environments. The same is true for flare systems, elevated pipe racks, derricks, and any other other structural assets as well. And with With Doc Drones, we can increase inspection frequency without increasing the labor requirement. We also reduce exposure working heights like we just mentioned and improve the documentation consistency. One example is that we created a program mission for the tank farm that executes a flight path that will capture specific tanks at set orientations and angles and it also labels photos to the tanks themselves so that the end user can identify changes through the consistent captures we take over time. The program tank flights are able to capture all tanks within one flight and provide data to the operations within an hour of us actually launching the drone. **[8:17]** So rather than replacing inspectors though, the goal is just to provide them with better information and so they can and do it more quicker and frequently and less and with less safety risk. The next is real the below example is we're looking at an image of a spring can. The spring cans mainly exist on elevated structures where they're providing an indication if there's thermal expansion on different piping systems. But the fact is that these cans usually are not very visible from the ground level and not and sometimes they're not even visible from the climbing platforms. So we've developed program flight missions that consistently capture spring cans on elevated structures and it's helped give us a way to visibly indicate where issues may exist on the structures and also limits personnel safety risk at climbing at you know to extreme heights. Next we'll take a look at security. Security use cases for dog drones provide a little bit of overlook value as I said before. Most refineries security systems rely on a combination of physical patrols, cameras, and lighting and and a growing network of sensors. **[9:20]** These systems work well, but they have their practical limit limitations. Our outside like I said before outside has miles of perimeter fencing, but not every six section of that perimeter has the same camera coverage. Not every location is easy for security personnel to reach quickly. So for our dog drones we're providing a way to those coverage gaps while extending the effectiveness and existence of existing security resources. We program missions that can scan the perimeter fencing in two flights providing video and images of all the fencing conditions, the sign is locations, and the gate positions security so security where they can examine all these areas of the perimeter. Um Uh and again like I said our objective is not to eliminate their human patrols. The objective is just to make the human patrols more effective by giving security actionable information before they arrive on site. scene. Uh and lastly we'll take a look at emergency response. Um this is possibly the easiest case for most refining teams to understand. Uh so during an incident **[10:21]** information may be becomes extremely valuable. The faster we understand what's happening the faster we can make a effective decision. Uh our remote flight operations can provide immediate situation awareness during fires, process upsets, releases, storm events, and in a multitude of other emergency scenarios. Uh rather than sending personnel directly into the uncertain conditions responders can make or can first gain an aerial perspective and understand what's going on. Uh this improves decision quality while potentially reducing personal exposure during the early stages of an incident. So before dog drones it could take our site up to an hour for a drone pilot to arrive on site, prepare the drone equipment, uh deploy to the scene, and then launch the drone for visibility. Uh but during several emergency response trainings with the dog drone we've been able to launch the drone and provide a live streaming to the emergency response team within minutes of notice of the event. So going from hours to minutes has been a critical change in the way that we're able to respond to these events. Uh so one question we've often asked is how do we identify how do we come about **[11:25]** identifying these use cases? And the answer is pretty simple. Uh we just spent time listening. So rather than beginning with the predefined list of drone solutions uh we've been meeting with departments across our refinery refinery and focusing on understanding that they're daily workflows. Just asking questions is like what task consume their time, when inspections are occurring repeatedly, where our personnel collecting information manually, where where observations critical to decision making and also like what what what could what could program drone capture reduce safety risk to people or where could it program drone capture. Once we began these conversations, the workflows sort of map themselves out and opportunities for Doc Drone became much more apparent. So these use cases largely emerge from the refinery operations themselves. Next I want to discuss a little bit about operating regulations for Doc Drones in a refinery. So our site obtained a BVLOS or as you may know visual beyond visual line of sight shielded waiver. We we successfully attained this waiver about two months **[12:28]** from the time of the original submission to the FAA. Our waiver specifically is a 200-ft critical infrastructure shielded waiver, which means we're able to fly up to 200 ft within critical infrastructure in areas unless we're within 200 ft of an elevated structure. Then we're allowed to fly 200 ft above that structure as long as we do not exceed the 400-ft hard ceiling. So looking at the map, critical infrastructure is defined as all area within the control access perimeter. Here you can see it's outlined in blue. So what that means is inside this perimeter, you know, we we we are able to exercise a 200-ft shielded waiver, but if we fly any paths outside of that, then that gets reduced down to 100 ft. So I also want to provide a few suggestions to anyone attempting to submit a waiver under similar circumstances. The first suggestion is to you want to set your limit to altitude about 200 ft or 60 m. Keeping This is because keeping operations at a lower altitude **[13:28]** reduces airspace conflicts and aligns with the standard low altitude infrastructure and of inspections. Also in your documentation and in practice, you want to try to stay close to shielded structures because operating near physical structures demonstrates that flights remain contained within a predictable control volume rather than wandering off into to open space. Uh Uh next one is you want to use and provide documented operating areas. So flying within established controlled environments provides the clear boundaries and limits on unpredictability for the FAA. Uh and one of the next points is of course is we want to minimize exposure to uninvolved people. Uh so both in your flight plans and in your documentation, uh you want to show where you're restricting flight paths and keeping away from the general public and any non-participants uh to ensure their safety, which is a primary regulatory concern anyways. And lastly, uh you want to present some sort of quantitative risk analysis. You want to be able to provide some sort of risk matrix assessment. Uh this could be **[14:31]** any kind of table that lists potential risks and your confidence in mitigating them. Uh so providing data-driven safety assessments helps prove that potential hazards have been thoroughly evaluated and mitigated, thus again clearing concern with the with the FAA. Uh another question we're usually asked is how do you fly and operate a BVLOS drone safely inside a refinery? The answer is that remote flight operations in these complex environments rely on layered safety controls. Uh so drone management platforms allow us to establish geofences, apply no-fly zones, set altitude limitations, uh set up predefined routes, and put with and with old automatic mission constraints. And our flight operations are highly structured and governed by these predefined procedures. So what you're here what you're seeing here is an example how we use these tools to create a controlled operating environment inside the facility. We have no fly zones applied to all areas of the site where there's structural or thermal hazards that **[15:32]** may exist at elevated heights. So even if a preliminary safety control like obstacle avoidance were to fail, these secondary controls provide a safety net of assurance for a smooth flight operations. We program fail-safe return home actions as well with the drone experiences a lost link or low battery percentage. The drone is preset to return to the dock safely out with while still avoiding these no fly zones and the return to home altitude is set is preset to a safe altitude or elevation. Lastly, we've also pre-grown pre-programmed missions that capture waypoints and return to home waypoints steer clear of assets. So our flight paths never fly over tanks and waypoints are set at least 100 ft from vertical structures. Even if those vertical structures are the ones that we're expecting, we still maintain a safe distance. So autonomous does not mean uncontrolled. In many ways, it just means it just means the opposite. Lastly, I wanted to just give a look ahead to our goals moving forward with our program. The first thing is we're looking to scale flight operations. So earlier we **[16:36]** looked at a few examples of program inspections that we set up for the tank farm and for Derrick structures. Next we're looking to increase our inspection missions to cover more structures using thermal sensor to detect thermal anomalies on assets like fin fan tubes and and heater stacks. These are two types of assets where it's important to detect that the heat flow and heat containment are occurring correctly. We inspect these assets today using manual control of the drone. But the next step is to move forward and create program inspection missions that allow for more consistent and frequent capture that makes it easier to detect the changes over time on these assets. And also in the security space, we're working on expanding the dock drone capabilities to assist security in our sites counter UAS response. Uh we're integrating our counter UAS detection system into the dog drone flight deck so that we can quickly launch the drone to a detection location provided by the counter UAS system and provide a quick on-scene visibility for security so we can help improve their response time in these with these kinds of events. Uh next we're also looking **[17:39]** to expand on AI detection capabilities and also establish automated workflows with the data captured from the dog drone. We want to provide insightful and actionable information without increasing burden on operations and security for the data review. Uh we can capture and provide valuable information from the tank farm flight but if tank operations has to review hundreds of photos to find issues then we haven't really created a great efficiency for them anyway. Uh so instead the goal is to utilize AI inspection tools offered by the dog drone platform to learn and spot where issues and anomalies exist with the data such as water on the tank roof and provide automated reports to operations or other stakeholders uh that this is in this example to indicate only the tanks where issues were found. Uh and also uh in in ROI or return on investment space where we're looking to to grow and expand on that. So from a management standpoint uh what's what's almost as important as the data captured provided to the end user are the numbers that support the value in the data captured being collected by the dog drone. So going forward **[18:43]** we're working with other department teams to establish accurate evaluations that we can derive from these program drone flights. So for example every time we complete a tank farm flight we want to accurately apply evaluation of exposure hour reductions of operations in the in the safety space while you know avoid them from climbing tanks potential financial impacts of any discoveries uh and any of any efficiency gain gathering information with the drone versus the time it takes to to do a manned inspection. Because ultimately, it's the value management sees from these missions that will help to grow and expand our doctrine program. Uh And lastly, we're going to work with to to for part 108 compliance integration. So, part 108 will very soon very soon require more detailed auditable aspects uh such as maintenance, operational records, safety management systems, training records, and risk management procedures. Uh the doctrine platform that we already use today offers many of these tools that will help that will help with accomplishing the **[19:46]** compliance requirements. So, we'll just continue to work with the doctrine management platform to ensure that there are no gaps in compliance when part 108 is released so that we can have a smooth transition with the with the new regulations. Uh now now I'd like to open up to uh any questions and discussion. >> Yeah, thank you. Thank you so much, Wade. Uh the way and the depth uh with which you have explained about the operations of this refinery are you know, actually amazing. Uh people never realize the the depth and the scope uh you know, until they see the map, until they actually see how many no-fly zones that you have to create, until they actually see uh how complex these operations are. So, one question that I had in my mind like you mentioned about pre-planning missions, pre-planning you know, programs. In such a complex space what type of like skill uh do you require to pre-plan these missions? Like how uh because again, there's a lot of different structures at a lot of different heights. **[20:48]** So, and running autonomous mission, running a remote mission, BVLOS, uh how do you create those missions with that efficiency? >> So, I think it first starts with a part 107 pilot who has plenty of flight experience even with manual control drones just to understand your capabilities and your limitations as they are as you approach different hazards and structures. And then obviously a good base knowledge of the site itself so that, you know, if you go do a survey of the site, you have a good understanding of the things you want to avoid and and and, you know, avoid that type of conflict and issue. >> Got it. Got it. Uh awesome. So, now I'll just share some questions, you know, over here. So, uh refineries run 24/7 and shutting down a unit for inspection is extremely expensive. How many inspections that used to require a shutdown, uh can you now do with a drone while the unit is still running? >> I'll say this, we we've replaced some inspections in some places, but uh for many of many of times what happens today is that the both inspections still incur uh and **[21:51]** now we're able to use the drone data to help validate or uh help them make a decision or something maybe they couldn't see. So, the the answer is that, you know, there there's a large and maintenance and inspection team and they still have a job to do and they still, you know, prefer to sometimes see a hands-on approach, but now the drone's more in the space of giving validation, like I said, visual visual confirmation of uh what's up there. And then some places where they can't reach, we're able to help extend uh some help there as well. >> Got it. Got it. Uh so, now we'll have the next question, uh which is how are you handling the data volume? Like a drone flying daily inspections across a refinery generates a massive amount of imagery. Uh are you are you reviewing all of it manually or is AI filtering? Like how How do you analyze the data, basically? >> So, right now, you know, the the the answer is that we have a sort of a heavy workload of collecting a lot of data and a lot of inspections and uh kind of sort of doing the uh filtering out ourselves manually, but we're like I mentioned before, we're sort of in a transition phase where we're trying to grow and and integrate **[22:51]** more AI stuff because, you know, that that's just the way of the future. Uh and I believe that's where efficiency will be realized and that's where the value we're going to bring to the end user because we're very quickly give them just issues that are necessary, not giving them again the whole multitude of data that they don't they don't need to see necessarily. >> Awesome. But in this case, like how do you go like how do you see the scaling? Like right now you mentioned that overall your operations will be scaling and again as you scale more data comes in. What's your team looking like? What your current team structure is in this case? So, yeah. >> So, we currently have 17 pods on site. Only about three of us are really capable of operating the you know on the beyond visual line of sight operations, but we intend to grow that in the future because for one you know one of the biggest reasons in emergency spots is having more people available more times that they can launch the drone so that you know one of us are all after business hours and not **[23:51]** capable of launching a drone. We don't want to have the gap occur there. So, the the idea is that as we grow further and we get more guys experienced with control flights that we would integrate them and get them available to fly beyond visual line of sight as well. >> Got it. Got it. Amazing. Now we have another question. When the drone captures an inspection image, who acts on it? Does it go to the reliability engineer, a maintenance planner, or the operations team? And how quickly does that happen? >> So, the answer to that really is all of the above or it could be any of the above. You know it just depends. We get requests from all three listed here and quickly you know right now again a lot of capture happens with manual flight. We tend to try to get a one to two day turnaround on any request. You know there's different levels and magnitudes of of capture that you need to accomplish sometimes inside of vessel, outside of vessel, but our goal is to strive to get it either the same day or the day following and it could be again to be any number of one of these departments. **[24:52]** >> Got it. In this case, if you were to divide this into uh percentages, uh what would the division look like? Like you get the most requests from which department? Is it security? Is it reliability? Is it operations? Is it tanks? >> Today I'd say it'd be the first two, reliability engineer group and and and maintenance. Uh the the maintenance and inspection guys really really uh they've really seen the value of drones and and yeah, and all kinds of elevated structures. Even if they don't necessarily have a way forward, they like for sometimes just to go try to see if we're able to get valuable information that they were not able to collect, you know, with scaffolding or ropes and and whatnot. >> Right. Understood. Uh now we have uh the last question uh from the audience. Uh you mentioned incident response. Can you walk through what actually happens when an alarm goes off uh or when an incident happens basically at the site? Uh how fast is the drone overhead compared to how long it used to take to get a person with eyes eyes on the **[25:53]** situation? >> So so the answer to this question is that it's really dependent on uh what time the incident occurs, right? Some some are more convenient than others. Uh if it's during business hours, you know, typically that means that me or any other qualified pilots with the drone here, uh as soon as we get the alarm, it could be it could be via phone or just, you know, an actual site-wide alarm. Uh in that case, you know, that's that's the most preferred obviously. You know, we're able to get the drone in the air and I'd say I mean not not more than two or three minutes uh with a location where we can go look for. We can launch the drone and and and you know, it's there with a a live stream available to our EOC and uh emergency response team. Now, go scenario that after hours, weekends, you know, there's not a lot of people here. Uh in that space, it's more dependent that I, you know, I'm just capable to to access the computer. So again, from our site lead or from some emergency response, I'll receive a call or text letting us know hey, something happened. I can jump on my computer remotely, you know, that's uh 60 mi away from the the get on my computer, uh launch the drone and the case again, as long as available, I'll able to get eyes **[26:56]** on it with less than 5 minutes. Compared to what the other question was is originally, you know, if we were we were all site the time and there was no pilot that were part one of seven pilot capable, uh it could take anywhere from 30 minutes an hour to actually just get to site. Once you're on site, you know, many times you need to start charging batteries, uh determining what drones you want to take out into the field, uh then begin packing, getting things staged, and getting out there. And then again, that that that easily turns into one, maybe two, three hours, uh depending on on what your your situation is. So, yeah, the response time has been greatly uh increased. >> So, with this, I have another question. Uh basically, you mentioned about battery charging and a lot of other things. Like these are some sort of hardware limitations. So, where do you think the current hardware is limited? Like with respect to the type of inspection you want to do? >> Well, I mean, not not to not to steer your your your your answer there, but I you know, batteries is is always usually the answer, right? Like the the amount of flight time you can get one flight to the amount you have. So, the you know, the if you have three flights, that may mean you only have 45 minutes of flight time, but **[27:58]** if compared to you have more, like the batteries always usually comes down to batteries, right? Cuz the especially with the drones we use today, uh their capabilities are are pretty impressive. >> Awesome. Uh we have less than a minute, so I just have one last question, you know, for the audience. On a large site like yours on a refinery, what are high frequency inspections that you have to do and what are low frequency inspections? >> So, I mean, the examples I use are the the high frequency. I guess it's the reason we're using it, right? So, everything pertaining to uh pertaining to the tank farm, uh coker derricks, like I said, those spring cans are very uh high up there and they're not very accessible or visible from from other areas. So, those are the ones that we we do in a bi-weekly to monthly basis. And then other ones sort of just range from there. Uh and they are more of usually they're ad hoc or one-offs when things are requested. >> Awesome. Awesome. Uh That is amazing. Wade, again, thank you so much for taking this session. I believe all the audience whoever joined, you know, gained a lot of insights on how to operate drones in a safer manner at all at a large site. So, **[29:03]** yeah, thanks again everyone and have a lovely day, lovely evening. >> Thanks everyone. Thanks for your time.