Become a Creator today!Start creating today - Share your story with the world!
Start for free
00:00:00
00:00:01
How AI simulators elevate feed mill operator training
56 Plays4 days ago
Ranjit Maharahjan, ANDRITZ head of product group - automation solutions, explores the use of digital twin technology to enhance feed mill safety, maintenance and operation.
Recommended
Transcript
Introduction and Sponsorship
00:00:07
Speaker
Hi everyone, welcome to Feed Strategy Podcast. I'm your host, Jackie Remke, Editor-in-Chief of Watts Feed Groups. This edition of Feed Strategy Podcast is brought to you by FeedStrategy.com.
00:00:21
Speaker
FeedStrategy.com is your source for the latest news and leading edge analysis of the global animal feed industry.
Introduction of Ranjit Maharajan
00:00:29
Speaker
Today's guest is Ranjit Maharajan.
00:00:33
Speaker
head of product group automation solutions at Andretts. And he recently spoke at the Feed Mill of the Future conference held at IPPE in January.
00:00:44
Speaker
The sold out event was organized by Feed Strategy and Feed and Grain magazines in partnership with the American Feed Industry Association.
Role of AI in Revolutionizing Training
00:00:53
Speaker
Today, Ranjit will talk a little bit about what he covered during that presentation specifically focusing on how Feedmills are revolutionizing operator training and decision making through simulation technology and AI-powered data tools.
00:01:10
Speaker
Hi, Ranjit, how are you today? Hey, Jackie, I'm doing fantastic. How are you? Great, thank you so much for being here. It's nice to talk to you again. Well, let's get right into it.
Challenges in Operator Training
00:01:21
Speaker
Feedmill operators are increasingly working with complex automation systems. What challenges does that create for training And how are operator training simulators being used to address safety and skill development?
00:01:35
Speaker
As the feed mill become more and more automated, the role of operators have changed considerably. It's not anymore just starting and stopping a machine.
00:01:46
Speaker
They have to manage a lot of interconnected systems. They have to understand the process behavior, and they have to respond to the situation when that's outside of their normal conditions.
00:01:57
Speaker
There is always going to be some upset conditions that is going to be thrown at you. And now today's feed mill operators have to basically figure out how quickly they can actually solve this
Benefits of Simulation-Based Training
00:02:10
Speaker
problem.
00:02:10
Speaker
So the challenge this creates for training is that many of these situations are rare, but critical. Like think about it, pilot, very rarely he's going to land on a water, but that training is necessary.
00:02:24
Speaker
You'll feel comfortable knowing that pilot has gone through that training, although the chances are very slim. Similarly, there is a lot of things happening in feed milk. There's a raw material variability, there's a process upset, or there is an equipment failure that just doesn't happen every single day.
00:02:42
Speaker
But when it happens, the operator needs to be prepared to really understand the problem and act on it quickly in a safe manner. So this is almost impossible in a live environment.
00:02:55
Speaker
Today, most of the training happens by work shadowing or by trial and error method. And there is a risk of safety, quality, and throughput when you are operating like that.
00:03:09
Speaker
So here comes the simulation-based training that addresses this and gives the operator a realistic environment, much like your flight simulator where the pilots are getting trained. we do have an option for the operators to sit in front of their operator station and they will not know that they are actually operating a digital plant.
00:03:31
Speaker
So this gives them the practice and responding to different situations. And we can actually program different scenarios depending on what typical challenges what they have or if there is something specific to your mill, we wanted to develop that, we could do that.
00:03:47
Speaker
Where then the operators are given these challenges, and they keep you know basically practicing. And what this does is that it actually gives them the confidence over time without really risking the production or safety of the female.
00:04:02
Speaker
What is powerful, what we have seen is that the operators kind of repeat this multiple scenarios, learn from their mistakes and build confidence. And when they are actually facing this situation in real time, they are very well prepared.
00:04:17
Speaker
So over time, This actually leads to more consistent decision making and much, much more stronger safety culture. So ultimately, there is a shifting in the training that happens from on-the-job training or trial and error to more of a simulation-based controlled environment, which is more structured, repeatable, and safe, which is exactly what other modern automated mills need today.
Conversational AI Tools for Efficiency
00:04:45
Speaker
During your presentation at the Feedmill of the Future Conference, you mentioned that operators can now chat with their data using ai tools. Please explain what that looks like in practice and how it changes the way Feedmill managers can make operational decisions.
00:05:02
Speaker
Absolutely. Yeah. So with the advancements of AI and LLMs, there is a lot of you know changes that's happening very quickly in a good way that actually fundamentally changes the way we actually interact with the data at the mill.
00:05:17
Speaker
So what in practice means chatting with data? It's really how people interact with the information. So traditionally, a feed mill, you have a set of alarms or your production report or your are daily report and people just take that when there is an alarm or at the end of the day they look at the production and they are more actually being reactive to it.
00:05:40
Speaker
Now what does AI chat with data does is really removes that friction where the operators and managers can simply ask in a plain language. I'm sure everybody is familiar with ChatGPTs and Copilot.
00:05:55
Speaker
It's basically the same ah technology that you would actually have in your feed mill. So for example, mill manager would say that, hey, I'm noticing that the pellet quality is slightly, or the throughput is slightly lower than normal, but there are no active alarms. So what is going on?
00:06:13
Speaker
So instead of like having ah an operator going and looking at it or a an engineer going and digging all the data and putting together the reports and trying to find the root cause, the system now does automatically for you. So it actually looks at all the historic data, it analyzes the patterns, and it actually digs through tons of data from your mill.
00:06:35
Speaker
And then it will say like, an example could be that, yes, your throughput has been decreasing by 4% over the last two shifts. And that is because it's correlating with the higher conditioner temperature variability. So you obviously have that upstream and then that is what is affecting. And also your energy consumption is actually a little bit more higher. And it might actually even go further and say there's a similar pattern in the past. We were associating with that dye wear some changes in the raw material moisture.
00:07:03
Speaker
So this level of insight is readily available at your fingertips. We're just asking these questions in a very conversational manner. So instead of reacting late or guessing, what this does is it gives them the confidence that they know exactly what the root cause is and they might be able to attack it. It could be something like you wanted to schedule a maintenance that inspection and then you wanted to act early before the quality gets impacted.
00:07:33
Speaker
So this fundamentally changes the decision-making at the mill so the teams can move much more agile And it's not like sitting at the, you know, the data is not sitting in front of like one or two people.
00:07:46
Speaker
What's important is this, the accessibility. What I love about is that the same data is available to the operators, to the mill manager, to the maintenance manager, and everybody is going to operate off of the same data.
00:07:59
Speaker
And it also actually levels the playing field because let's say there is an experienced operator who over years of experience can understand this pattern. But when there is a new operator, then they're going to be completely lost.
00:08:12
Speaker
Now he would have the same access to this same data and he would be able to take a much more informed decision. So what I have seen is that over time, this changes the culture.
00:08:24
Speaker
So the teams actually starts asking better questions. Even when I started asking questions on ChatGPT, my prompts are getting better and better. You know how to navigate through that. So similarly, you get better and better in terms of asking better questions. And really, then the AI becomes your decision-making path.
00:08:42
Speaker
So basically, you're having ah really one of your best operator with all the data insights sitting next to you. And that's what really this you know chat with data is going to bring to the feed now. Then just to clarify, this is something that mills are doing right now, and it's not something that's available down the road?
00:09:00
Speaker
No, it is something that is available. And we have actually, this technology has been there for other process industries for quite some time. So this is a very proven and vetted technology. For now, it is already available. If somebody wants to explore that and they wanted to experiment, we definitely can help them out.
Addressing AI Misconceptions
00:09:20
Speaker
Based on your experience, what are the common misconceptions with AI and what results have you seen when companies do take the leap and trust the technology? i think this is an excellent question because I actually asked about this a lot of time from my customers because i think There is the word AI being thrown at everything.
00:09:39
Speaker
So sometimes the customers are not really understanding it. So the biggest misconception consumption is that they feel like either the system is too complex or it's on the other end of the spectrum. They might think that this is all about like just dashboarding.
00:09:54
Speaker
So there is a lot of confusion. And also, and this is even like from my real world experience, the operators are actually scared. There was a moment when I remember when I was commissioning the system,
00:10:06
Speaker
and the operator asked me point blank is that, hey, will I be fired after this system is in place? Because there is a fundamentally, they are scared about this technology as well. So many teams worry about introducing this AI, and there is a lot of confusion about it. And also they worry about the the effect of really good operators or experienced operator.
00:10:28
Speaker
In reality, it's actually quite opposite because What happens is that the AI is actually enhancing the operator knowledge and also helping them along the way to be able to make better decision making.
00:10:42
Speaker
So when applied to real operation problems, it actually amplifies the human expertise. It doesn't replace the humans. It actually amplifies that and produce tangible results.
00:10:54
Speaker
I can give one example where recently there was a mill and we had vibration sensors put it in there. And in feed industry, we always don't have the luxury to actually analyze all the vibration patterns. We don't have a dedicated vibration experts and so on and so forth.
00:11:11
Speaker
So we wait until there is alarms or something like that, which is already too late. So when we implemented our system, the AI detected actually a pattern, which was what was happening before.
00:11:24
Speaker
And it actually gave an early sign, even though it was well within the limits. we were able to predict that if this continues, then you're going to have a bearing problem.
00:11:34
Speaker
And so the the system actually alerted the operator saying that, hey, the AI has detected some patterns and then there's early signs of bearing looseness and you might want it to do some corrective action.
00:11:45
Speaker
So then this actually let the operators to kind of like basically finish the production, slow down and do a maintenance check. And they were literally able to say two days of downtime.
00:11:56
Speaker
Because if it had failed, then you know that would have been a much bigger disaster. Another example comes from a production throughput and quality. So there was a subtle deviation, and it's almost impossible to track that deviation. And that AI looks at not only your current production, but it actually looks at all the other parameters, your minimum sustainable rate, you know your overall equipment efficiency. So it looks at a snapshot of your current system because it's not always optimum. There things that are you know not optimal.
00:12:25
Speaker
And it takes into effect all of these things. And then it will say that, hey, there is a bottleneck in the conditioning, for example. And the operators proactively would adjust this team and then be able to optimize that rather than that particular batch becoming a scrap or something like that.
00:12:41
Speaker
So what I have seen is that the most significant impact is the decision quality and the team confidence. So once this team actually sees the AI can help them with the real problems and solve the real questions in their head, then it actually becomes a partner and not anymore seen as a threat or a confusion.
00:13:00
Speaker
So the skepticism drops and the adoption grows, but there is a curve that everybody has to go through. And, you know, the focus really shifts from fear or doubting this technology to be using this tool to help making a faster and safer and consistent decisions over time.
00:13:18
Speaker
You made some good points there. And I think highlighting those practical applications within production is extremely important at this point because it may be difficult for operators to imagine without that context.
Successful AI Implementation and Cultural Impact
00:13:30
Speaker
Now, are there any challenges to implementation for feed mills that are interested in exploring this kind of integration? Yeah. So typically, what I always say is that this is a journey.
00:13:44
Speaker
There are a lot of solutions that are available. Typically, how we operate is that We go and listen to the customers, what their challenges are and what their main issues are.
00:13:54
Speaker
And then based on the experience and what we have, our feed expertise knowledge and the the tools and everything, what we have in our tool set, we would work with them and really put together a solution.
00:14:06
Speaker
So this is not like a product that you just go and install and disappear. It is something that you have to collaborate with the females. You have to work with them and understand if their pain points. and ah bring the right technology and the right solution.
00:14:19
Speaker
Because one of the risks is that if you are not understanding the pain points, or if you were to actually put in the wrong solution, it just becomes an expensive venture and everybody's going to lose confidence. So we have to be extremely careful in understanding what is the challenge that we are facing and how do we solve it? And do we have technology that can actually solve it?
00:14:41
Speaker
So if you were to ask, like even just a year ago, there were like a lot of technology that ah has completely shifted the way we are working. And so as the technology grows, the solution grows as well. What I would suggest is that even if somebody has looked into this technology before, it might be actually not a bad idea to look into it again, just because the technology has been growing very rapidly.
00:15:03
Speaker
And when you reflect on the feed mills that you've worked with, the ones that are really embracing this new technology, what would you say is the most surprising transformation that you've observed?
00:15:16
Speaker
So what I have personally seen is that it kind of starts with a very low confidence and even doubts to some extent. And then once they see the results and once they see the effects of it, they understand, ah, this is actually helping me and not trying to replace. That transformation is what's really surprising.
00:15:36
Speaker
And you look at it like there is about roughly 20,000 feed mills around the world, just in North America, 5,000 to 6,000. And on average, if you take about six or seven operators, and to some estimates, we have about 15% attrition rate. Now, you're talking about close to like, you know, 28,000, 29,000 people that are churning every year.
00:15:56
Speaker
We have to understand the challenge here. I mean, the scale of the challenge is huge because you simply cannot just do an on-the-job training or work shadowing or just a trial and error with this amount of churn that is happening. And in addition to that, it's also sometimes very challenging to get operators in the feed mills because of, you know, sometimes the location and everything. So that's why it becomes really important that we are offering this simulation-based training where
00:16:29
Speaker
You actually can practice in a virtual environment, in a very safe environment. And the AI system not only just helps you with the training, it also makes you a better operator. And over time, what I've seen is that operator gets complete confidence in it.
00:16:45
Speaker
And people that were like really detractors actually started supporting it. And they even want to expand into other areas of the plant. So what's remarkable in one word would be the cultural transformation.
00:16:58
Speaker
So the operators are able to collaborate more effectively and share insights and trust both their training and the recommendations that the AI is giving. So they see that their overall skill level is also coming up and what it used to take years and years to get to that level of competency. Now you're able to achieve it in a much, much faster.
00:17:20
Speaker
And even sometimes like, you know, a few days or months, you would be able to get to that level. So it elevates the overall team performance and resulting in a a much more smarter, safer meld.
00:17:32
Speaker
Excellent. Well, thank you so much for your time today, Ranjit. And thanks to our audience for tuning in.