2025 - AVEVA World - San Francisco - Engineering
AP Consultoria: Digital Transformation in Engineering - AVEVA UE and AI for Multidisciplinary Projects
The case study clearly shows how using AVEVA CONNECT and Unified Engineering delivered cost savings, increased efficiency and improved collaboration in the design of the new gas outlet in Amazonas. The use of 3D modeling, process simulation and integration with automation ensured greater reliability and agility in execution. The integration of AVEVA Unified Engineering with Artificial Intelligence marks a significant advancement in multidisciplinary engineering, bringing greater efficiency, precision and automation to industrial projects. The connected approach between diagrams, 3D modeling and engineering data reduced errors, optimized collaboration between teams and ensured consistency at all stages of the project. Additionally, implementing AI for pipe support prediction demonstrated how the use of machine learning can transform processes, reducing modeling time and ensuring smarter, more reliable design. With this innovation, the engineering sector is moving closer and closer to an automated, integrated and data-driven approach, where decisions are based on predictive analysis and not just manual experience. This project not only validates the effectiveness of these technologies, but also paves the way for new applications of AI and AVEVA Unified Engineering in engineering.
Industry
EPC
Company
AP CONSULTORIA E PROJETOS
Speaker
THOMAZ COSTA
Session Code
SESS-212
Transcript
Hello there. My name is Thomas, and it's a real pleasure to be here alongside with a great friend of mine, Tiago. And we will share we're excited to share our the experience we have faced and the solution that we have found integrating the AVEVA unified engineering and AI solutions on the multidisciplinary projects.
So I hope you find this insightful and also enjoyable.
So here's the basic table of content we're going to share. We're going to pass through the objective, our goals about the AP, the unified engineering solutions, the change that we have made in the last two years. We used to use the license on premise, and then we have changed about the project, the case study. And then Thiago will dive into the AI case study we have made it.
So this is just the summer we have made integrating AVEVA and engineering and AI. So the main challenge you have faced is how could we reduce the hardware and software costs, infrastructure's costs?
How can we facilitate the integration between all the teams and the disciplines?
How can we have the project quality and the agility enhance it? And how could we introduce and have some in in ETAs of it initiates of this on the on the AI on our company? So the solutions was basic integrating AVEVA, unified engineering, and got some students, some people that really had hard work on trying to implement AI and integrating these tools. So the results was amazing. So we have more than fifty five percent, fifty three percent in saving converting PCM, the Point Cloud Manager softwares, and fifty six percent in reduction hardware cost, six percent faster documentation, and twenty percent less rework. When I'm talking about fast documentation, it is the speed that we are issuing documentation.
Thirty percent, thirty four percent hardware cost reduction, and fifty three time percent saves in converting to E3D. And this is the number that for me is more amazing. That plus ninety percent reduction in the time spent analyzing the type and position of the support that we do basically today manually. So we have many hours.
He looked into support. He took a table. And he think where, by some math equations where we have to put one, two, three things, and now we have really more than ninety percent of reduction. So sixty percent reduction of time stress analysis review and twenty percent reduction in support modeling time.
So these numbers for me is really amazing. We're going to show you a little bit about this forward.
So let me briefly introduce myself. My name is Thomas. I am a civil engineer, and I've been working on the engineering design company for more than seven years.
And also, I'm a CEO of the company. And also, am as a volunteer work, I am the managing director of Brazil Association of Industrial Engineering and Design. Please, Thiago.
Sure. Thank you, Thomas. I'm Thiago Rivera. Many of you know me already, but I'm director of tech support for AVEVA, responsible for the engineering and simulation software for the Americas region. And we are working together side by side with some customers in this AI initiative alongside with the support activities that we also deliver.
So a little bit about our company.
We are headquartered in northeast of Brazil.
We are an engineering design company with a wide spectrum domain, including pipe, instrumentation, electrical, chemical, process, and all the disciplines that are needed to build or design a plant or revamp it. So we were founded in nineteen ninety three. We have more than thirty years of experience.
One of the main points I'd like to share, we have been elected we were elected for three times in Brazil, twenty twenty one and twenty twenty three and twenty twenty four, one of the most innovative companies on the country. The main reason of this was basically about using AVEVA tools, integrating with laser scanning, PCM, and all of these technologies that we're going to share with you. And our partnership with AVEVA had more than twelve years. We have started it in twenty thirteen, and we keep going and we keep growing this partnership. So let me introduce you first the AVEVA Unified Engineering, the change that we have made on our company.
Until twenty twenty three, we used to use our license on premises, and the AVEVA team introduced it to us the opportunity to use this platform as software as a services and the metafeet of it. And now I'm going to share just some points. It's eight points. I'm going to pass a little bit fast to it, but you're to understand each one of these points. So the main point, one of the main points for me is the technical support. As you know, AVEVA license, we by we can pay by day, but we've been we use also by month.
And we have, like, eight hours work hours a day, but also we can use at night. So to optimize the use of it, sometimes we work also at night with our employees.
In the AVEVA engineering solution, we have twenty four hours, seven days a week technical support.
And also, we have less dependency of our IT team.
The secure and the way that we work now is really more safe and secure. We have more access user controls. We can track who are using each time, when, the time that the people or the person, the employee had used that license, and also the backup, considering that all our data are in the cloud.
Reducing of investment for me is another of one of our main points.
We used to use our hard, expensive machines and equipment, I seven, I nine, thirty more than sixty four gigabytes of memory and and all the other stuff we had to buy to use on premise. But now, considering that we are working totally on the cloud, we just have to have a fast internet and a fast connection to work and to use all the softwares.
Again, consider this. We have less IT dependency on our company.
Information are shared more easily, and we have greater transparency for customers. The real time project monitoring, we can collaborate in an easily way so we don't have to update manually with administration employee that was, at the end of the day, doing his work just to update staff and connection. Now the the connection, the collaborator are more are are easily and faster than it used to be when we had the license on premises. Also, the remote asset in platform. This is, for me, it's other real benefit because with the home office after the pandemic period, we have to use some VPN, some Citrix, and now we can work totally with just the assessing our employees on their home.
And all the softwares that we have to use, we can use it all on the in the cloud. So we don't have to buy the license separately. We just have to take the tokens, and we the tokens will be consumed by the using that we are going to work. And so we use now in our company AVEVA Engineering in diagrams. So the integration of data and sharing in real time collaborations are easily and more fast. Also, Point Cloud Manager now are online. We're going to show you a little bit some demonstration of it to you.
The process simulation that we use, now we have integration, and we have almost sixty percent faster documentation that are being issued to our client. And we can use it on the that we are working with the AVEVA defined engineering, and also the less we work, considering that we are integrated with other softwares.
The three d design, for me, the main thing of the three d design is the less cost and the cost reduction with the hardware, the high expensive equipment, and also the less rework that we had using in Unified Engineering.
Here is our basic workflow when we work, especially on brownfield plates. So we go to the field doing a field survey with the laser scanning equipment. We upload it to a VivaPoint Cloud Manager.
And depends on each case, we can use the Caesar two to modeling and doing the pipe stress analysis. Or in parallel, we can do the two d softwares, that is AVEVA diagrams, AVEVA process simulation to do the mass and energy balance, and also electrical instrumentation.
And then finally, we can integrate all of them on the AVEVA engineering into export documents, data sheets, and bill of materials.
So I'm going to dive a little bit on our case study. So the object objective of this case study was to implementing a new natural gas outlet in Amazonas. So it's a was a new thermoelectric plant with eight hundred and twenty five thousand cubic meters a day. So it was a multidisciplinary project with more than six thousand square meters scanned. The project time was on nine months.
And we had two fifty nine documents issued on that time.
So this is I'm really proud to show you this from me this year because two years ago, we were here. And I was just showing the integration between the AVEVA Point Cloud Manager and also the three d using we didn't use it that that in that case, in that time, the the AVEVA Connect was totally on premise.
And now we had the AVEVA diagrams that was the the PN IDs and the PNIs. The PN IDs totally on the AVEVA diagrams uploaded and built on it.
The AVEVA process simulation doing the balance of mass and mass and energy balance, AVEVA a three d design, and also the point cloud manager the the point cloud manager that we have uploaded our point cloud.
This is a video of the point cloud manager. So here, we can see a plot plan. We can have the bubble views. In the bubble views, we can see the picture. There is four ks. And we can see the modeling also at the same time of the picture that we have modeled for this project.
We can also see a particular equipment. We can take measures with not more than one or two millimeters of real data.
And also, we can go walk through it. This is the laser scan in the point cloud.
As you can see on the left side, we can tag the names of the equipments, the piping, and we can do whatever we want. It's like a social media that we can have on our company talking about Point Cloud. We can deliver to our clients a kind of a Google Street view to him with just a local access, and he can see whatever he wants to see in his project.
So this is the the final case. What what this picture showed to us is that I have a project that are with a great quality, that are integrated, And probably nine hundred percent, we don't have a problem of tag. So we have a tag on P and I. We have a different tag on data sheet. So when we integrate all of this information on AVEVA engineering, we ensure to our clients and to our team that all the design and the data are integrated in one page. So on the left, you can see the diagrams.
On the middle, you can see the model three d that we're designing on the three d. And on the right, you can see the data sheet that is the 1D document that we can have the pressure, temperature, the flow rate, everything we had of information of that equipment or pipeline in something like that.
So now I'll give Thiago to take over All right.
And explain a little bit about AI.
Hello, everybody. And thanks, Tomas, again, for this insightful presentation.
Actually, it set the stage perfectly for the second piece, which, by the way, I'm not sure if you're aware of this or not. But this will be the first case in AVEVA history of a customer presenting a real and complex case using machine learning and e three d together. So please join me in a round of applause for AP.
My mission now is to explain how AVEVA and AP work together to deliver this crazy solution of creating a pipe support prediction tool in a very accurate way.
And my plan here is to show you the different steps we took, including the preparation steps as well. And this is here where you start, explaining that before you start developing, before you start calling your data scientists, your experts, you need to check and map the challenge that you want to address, the problems that you want to solve.
And my goal here is not to explain the challenges in supporting pipelines, because I'm pretty sure you know this already. But the goal here is to show you the importance of knowing the challenge that you have. Because by doing that, you will be able to drive the development of AI solutions in a very more accurate way. So here we are talking about time spent in the process. We are talking about using expertise, knowledge from previous engineers and from previous projects. And we are talking about interacting with different disciplines as well, which I will explain the importance of having other disciplines involved in development of AI tools.
Because sometimes we are not talking about only AI tools, but also hybrid solutions. And I will explain a little bit more about that in further slides.
And after you map all of your challenges, now it's time to review your existing workflow. Because when you review your workflow, you map the steps that you have. But most importantly, you map also the data points that you use in each step.
And why this is important? Because knowing the data that you use normally, you will list all of this data and use it to train the machine. You will get only the relevant data and select what is important on each one of these steps and provide the machine these data points. Meaning that you should not be providing data that you don't normally use unless data scientists figure out that this is a very important piece of the work.
And it doesn't matter also if we're talking about software that are not within AVEVA portfolio. In this case, are talking about ProPipe, AP uses for preliminary determination for support, and also Cesar Chew. So a very important piece of this work is that after we used all of the inputs that Cesar uses for stress analysis, we figured out that in the end, the machine learning tool was able to kind of predict what Caesar would do, reducing the need of using Caesar too much, which is also one of the benefits AP realized at the end of this POC. So instead of having several rounds of stress analysis reviews, they reduced that significantly.
Because now A3D was able to provide them a first great gas for the support sub location and quantities.
Now that you have both challenges addressed and workflow mapped, and with that you have the data that you normally use as part of this process, it's time to develop the tool. And in this case, we separated this development into four different buckets, which is the first one, as I said, one of the most important ones, is defining your input data for the machine. So you need to be sure that you are using relevant information and you're using accurate information from previous projects.
And you should also consider using some data scientists and people knowing that the data that you're using is excluding outliers. So you're not affecting the way the machine is processing the data.
The second one is about revision history. So this related to the amount of data that you have available for the machine to train. Because I know many customers who know their processes, they know their challenges, but they don't have enough data.
And this leads, let's say, having the machine not getting the full picture of your process, of your problems. And as a result of this, the machine learning process cannot give you an accurate answer.
So make sure you have thousands or even dozens of thousands of data points when you consider creating machine learning models like this.
And once you train the machine, now it's time to validate the outputs. And here, after we delivered the first preliminary results from the tool, AP and our engineers validated the initial output for the tool.
And they were very, very happy with the initial responses. And from that, it's time to enter with what I'm calling a hybrid solution. Because you know that when you're supporting pipelines, we also need to consider other elements from other disciplines, like structural profiles, for example. So we cannot rely one hundred percent in AI responses, because you need to also let's say, if AI says that you need to put the support here, but you have a structure profile there, we need to adjust the position of the support to match the structure profile. So in this case, we are using a hybrid approach, is using the initial response from AI plus some design rule based mechanisms to refine the output from AI and provide almost one hundred percent accurate response for AP and their engineers.
And so this is what I'm calling a hybrid approach. And this is basically what most of the companies working this AI area are doing nowadays.
And just to show you that things might be simple from a user perspective who are just clicking a button to create a support and that's it, or even a preliminary support or a detailed support, in the background, depending on your requirements, depending on how far you would like to go and which kind of information that the AI tools should be able to predict, you need to multiply that by the number of prediction that you would like to have. So in this case, AP came to us and said, hey, AVEVA, I would like to do three kinds of predictions. The first one, I would like to predict the number of supports in the pipeline.
The second one, I would like to also predict the location of supports.
And we said, mm Okay, keep going. And then they started saying, Okay, not only this, but also the type of the supports. They'll say, hang on. Hang on. Let's let's stop this. Right? Let's stop here for a moment then before we can continue this very big challenge.
And guess what? After two weeks of work of two people, like I'm two great people, but two people working in this project, we were able to deliver to AP the first version of the preliminary support that we are doing as a POC to create preliminary supports. And these, I think, are the results that they got from this preliminary version. Right? So they already realized a lot of cost reduction in project execution time. They also increase the reliability of the project data and accuracy as well, because we are using this hybrid approach of mixing AI output and design rule based adaptation.
And also with more integration with other disciplines. So less discussions with disciplines around changes, because they were less frequent as well.
And here is one of the most impactful slides. And again, it's using just the first preliminary version of the tool, which is only creating the preliminary supports. But the video that I'm going to show you here is going to present the detailed support as well.
They are showing twenty percent in reduction in support modeling, thirty five percent in reduction project time, thirty five percent. That's a lot.
Sixty percent in stress analysis review, And this incredible ninety percent and plus, right, in reduction of time spent in reviewing supports, the position, and the type, which is impressive.
And to give you a little bit of a more tangible number, and here, with this example they ran within their systems, they created all of the support for seventy four pipes in just four minutes. And in this case, they created just the preliminary support because that was the version they had at that time.
But the video that have here is actually showing you the latest version of this prototype, which is creating also the detailed version of the support using the support's application library that we configured specifically for this project.
So here, you can ignore this interface because the interface is just pure PML. But it's using the so I mean, it isn't giving me the time to say that it created the supports, the preliminary ones, in a couple of seconds. And I'd like just to show you this example here, right? Because you see one ancillary support that is not matching structure profile, which is a good example, in my opinion, because it shows that the solution is not perfect. So it'll miss in some of the cases. But still, it will help you a lot with addressing to almost all of the other supports that you have in the pipeline.
And while I'm talking, now the tool is creating the detailed supports using both information from AI as well, but also the library that is available for the supports application. So here you will see that we are talking about an actual supo element from the support application. And here, since this pipe was previously supported by AP using at element, we were able to compare the the adder elements, with the new super elements that we added, confirming that all of the supports that were previously added to these pipelines were matching with the new ones using AI technology. And just to give you one example, we are talking about twenty five seconds only to support the pipes, to twelve supports for this pipe, the preliminary ones, and around forty six seconds for the detailed version.
So you see here that just two people, two great engineers, supporting working with a group of engineers from AP, providing insights, providing their experience, providing data, which is very important, were able to develop this kind of tool that will be massively helpful for AP in their future projects. So if you are also interested in knowing more about this, please reach out to Thomas, reach out to Marcus, who is in the back, or reach out to AVEVA, because we are always here to help you. So with that, I would like to thank you all and open the floor for any questions you might have.
