Industrial machinery and AI transcript-episode 2
Chris Pennington: Welcome back to our conversation on AI in industrial machinery. In today’s episode, we will be continuing our conversation with Rahul Garg, Vice President for industrial machinery vertical software strategy, and Shirish More, Product Manager – NX Architecture, AI & PLM SaaS Transformation.
Rahul Garg: You know, some simple examples I can think of, for example, material properties or designing for manufacturing needs, trying to do a lot of that analysis right up front during the design process itself. Those are typically a lot of capabilities that are in some cases maybe industrial insights that we can provide like material properties and information.
But the design for manufacturability is maybe perhaps a company process right on and how they would want to manufacture something, how they would want to assemble something. Bringing both those things together would make it a lot easier for the design engineer to make sure that he’s accounted for all those requirements right up front in the design process itself.
Shirish More: Exactly, exactly. And this is where you know I’ll introduce a term called industrial foundational model. And it’s the model that we’ll develop but it will have the ability to also get trained on customers IP and this model is purpose built to understand the language of engineering and manufacturing. And unlike the general-purpose models that are out there, these models that we will provide our customers will be trained on structured and unstructured industrial data such as CAD, geometry, simulation files, and to your point, validation data that they might have from their labs or some other places.
And once these models are trained, these models are then context aware, meaning that they can understand the engineering intent, recognize the geometry in context to which the user is asking help, interpret the process data, for example, resolve the ambiguity and then is going to start helping the users tackle the real-world industrial scenarios. That’s the plan and absolutely Rahul, you got it as to where we’re not just any AI model provider, we are here to help our customers leverage AI fully to speed up their overall product design processes.
Rahul Garg: You know, one of the other examples that comes to my mind is always around this area of MBD, which is, which is not exactly the most desirable process many design engineers want to engage in, because that requires them to go through a lot of detailed documentation on what providing out of PMID data right for manufacturing purposes, but now with using AI, all of that can be self-generated as well.
Shirish More: Right, right. And I think we are more than happy to engage with customers. In fact, I have a number of customers who are looking at both the 2D drawings and to your point the PMI wherein they have approached us, you know in a way in which like you know they have mentioned a couple of times to me that they have 20 years of worth of data, 2D drawings, blueprints, and whatnot. And it’s a perfect use case wherein we can apply the AI techniques, learn from the knowledge that’s hidden inside of these drawings and PMI.
And then next time when the user is in a similar situation wherein, for example, you want to reuse the data or reuse some of the things for a newer program, then you don’t want to start from scratch. These models will be able to then automate, start predicting and suggesting, for example, in 2D drawings, what are my standard nodes? Are you working against an assembly drawing?
Notes that you need to be aware of.
If you are working in context of a PMI, where should be your data planes and other things just by looking at the shape and other things. But you get the point as to having the insight into your best practices and historical data. We can easily reach to a point where next time when you are working on something similar. We can auto complete, auto populate and predict annotations instead of 2D drawings and PMI. OK, that’s the beauty of AI.
Rahul Garg: Yeah. And one of the areas right from an industrial machining perspective, I know many of our customers still rely on 2D drawings once the initial design is done to send it out to the shop floor, right. And I’m just wondering given the capabilities that are going to be enabled through the through these processes. Will that perhaps give them some impetus to try and maybe move away from 2D drawings? Any thoughts on that?
Shirish More: Yeah, you know to your point there are still processes which will need 2D drawings, . Like downstream processes, especially the machine store or guys who are on the on the shop floor. Sometimes they prefer having a 2D drawing there in front of it to measure, you know, cross sections or whatnot. But to your point, yes, if the AI model is trained, regardless of whether the user is interested in a 2D drawing or 3D PMI, because from an AI perspective it will be trained using the entities like the edges, the holes, and the manufacturing features so the automation as to where to apply that comes later on.
But if we train it right then I don’t see any reason why the customers might just ask the AI engine to start populating the 3D geometry with the PMI information. I see them once they gain the confidence as to the AI is basically predicting a lot of things in the right way. They might just ask the AI to generate PMI and then if there are downstream processes, we still need 2D. The same AI can easily generate a 2D drawing for downstream processes. And that’s where I see the world heading towards. Let the AI do it and based on the stakeholders, it can just generate the content regardless for 3D or 2D, it should really not matter, OK.
Rahul Garg: And I know we have also got our Cam assist for NX, which will allow you to basically create some machining process and the tooling and manufacturing constraints, right to validate and finalize the tool paths as well. I would imagine just those are like the next generation and supercharge your adoption of AI.
Shirish More: Exactly, exactly. I think once we and again I’m going back to our discussion around how we will provide customers with the ability to learn from their historical data. And when I say historical data, it may not be only the design data. Anything that’s associated with those design data, to your point, it can be features that are going to get CAM operations.
If we have access to the CAM operations, we can easily learn from the features and the associated operations and then auto generate the toolpaths based on what the AI engine might have learned. traditionally users might have made a mistake like they might have gone ahead and picked a tool with the radius 5. I’m just giving an example wherein the design might have used a minimum radius of something else.
And now with AI you can easily learn from previous designs and associated tool path. Now you reach a point where the manufacturing engineer is going to say that you can generate a rough toolpath for me and he’s just going to validate it. But that’s where you know that’s where the world is heading is. As long as we have access to some proven historical data both from design and manufacturing, we should be able to learn from it, start predicting things and guiding user to just speed up the overall process.
Chris Pennington: Which of our customers have used Design Center to their advantage and what benefits did they reap?
Shirish More: Yeah, that’s a very good question. And before I get into the different customers, customer types and other things, I first would like to touch at a high level some of the challenges of bringing AI into our mechanical product designing kind of industry, data privacy and IP protection majority.
When I talk with the customers, engineering data is proprietary. Making sure not only the engineering data but also machine learning model and deployment and inferencing based on the context in which the user is asking question, this whole idea about security or discussion around security comes up. That’s one thing. Ensuring that the AI decisions are explainable and traceable to the engineering rationale or standards is key to the adoption. I always tell our customers that we are not trying to replace the designers or engineers. Humans will be always in loop.
So, how can we make sure that the AI decisions or predictions that the AI models are doing can be trusted? Engineers, designers must trust the suggestions and predictions which requires some nicely domain trained models and verified outcomes. Keeping these key points in mind and as well as regulatory compliance and safety, especially when I talk with customers from aerospace or medical devices, they want to make sure that the AI models themselves are they certified and are audited.
Now that you see that whenever I talk with different customers on AI adoption, these are some of the key points that come in mind. Keeping these key points in mind, what we have started doing is, as I explained earlier, we started rolling out some generic AI models which are trained using some generic data like for example selection and clustering algorithms that can detect shape and other things which are not proprietary, which are generic. But this gives us a way into deploying these at different customer locations getting their trust and then expanding onto our different phases that we talked about.
Keeping all these points in mind, some of the customers who have already started adopting the solutions that we have delivered, especially in sort of Design Center, our customers within the automotive industries, a number of customers, both the OEMs and suppliers. I like the way we are heading because first from a deployment perspective, some of these models are deployed within their premises and once I show them the architecture, they are so happy because none of their data is leaving their premises. These models are deployed within their premises, the data that gets used for prediction purposes and suggestions as part of the Copilot architecture stays within their premises and they can see how it’s helping their users.
Automotive is one of those industries who has started adopting AI at a large scale, both the OEMs as well as suppliers. The electronics and semiconductors are another key industry wherein they really like what we are doing, especially now that we have rolled out Design Center Copilot and ability for Copilot to start suggesting things based on the prompt is another industry which is really seeing a value.
Going back to Rahul’s question around how you can look at the user’s profile and based on the prompt he’s asking, for example, when I get an assistance from Copilot, based on the prompt, I can easily figure out whether the user needs help. Maybe he’s listening. From our competitors’ product to Design Center and then based on the prompt the Copilot helps.
We have a number of industries especially in electronics and semiconductors who might have transitioned over from our competitor products to Design Center and now look at Copilot to quickly onboard that user onto Design Center and help him understand how to find an equivalent command in Design Center. This is where that particular industry is really adopting some of the Copilot capabilities.
And then finally keeping the electronics and semiconductors we got other customers who are into designing standard parts or majority of their products are very similar but are driven using industry-based practices, rules or expressions and they see how AI can like for example the connectors or you know there are a number of other contact points and whatnot. And these are day in, day out, same metadata that the designer needs to be aware of.
This is where AI is really helping them because it can learn from their existing data, it can learn from the shape and then start making the mapping between the metadata and then the shape. And this is another area where IC industries are adopting AI quickly because these are some of the repetitive tasks and things that the designer might normally miss out.
For example, applying a metadata or parameter onto a contact point and with AI in background, it’s looking at the context in which the user is applying or using a command and can start guiding the users. You know electronics and semiconductor, some of the medical device’s industries and others are now jumping onto this this AI world and starting to adopt our solutions.
Rahul Garg: Yeah, yeah. In fact, you know the, the one example you gave about connectors, to me a lot of those are like configured parts and configured products and typically in industrial machinery kind of applications. That’s one of the big parts of the business engineer to order and configure to order equipment with a lot of repetitive designs, but some configured changes.
And I think AI can help tremendously in those situations. I did have one more question though. Shirish, on some of the examples you spoke about. How do you see AI being adopted by medium sized companies, small sized companies as well? Or is this something that’s going to be limited to large enterprise organizations?
Shirish More: No, and that’s a very good question because, well, the medium size or very small customers don’t have the luxury of hiring data scientists and AI engineers to look at their structured data or unstructured data and historical data. They are going to heavily rely on the Siemens tools that we will deliver with our out-of-box solution offerings.
In fact, I see medium to small businesses really leveraging the power of generic models that we release, models that are trained using Siemens IP model that are trained using our best practices, our application knowledge and these customers will in fact adopt our out of box generic models to AI engineers. They are relying solely on our expertise and providing them with generic models. In fact, you will see small to medium businesses adopting out-of-box AI in a much faster way as compared to the large enterprise customers.
Chris Pennington: Do either of you have any closing comments or thoughts? Where do you see AI going in the upcoming years?
Shirish More: Well, I’ll start and then I’ll hand over to Rahul. Clearly, based on today’s discussion, AI is not a hype, . Means what excite me most is how AI is becoming genuinely useful. We talked on some nice concept and those are no longer concept because we have already started implementing it as part of our Design Center architecture.
Things like the agentic workflows or domain specific Copilots, you know we are building these tools, which will solve real problems for engineers and manufacturers and in form of our digital industries, AI model, foundational models. We are making sure that the AI that we will deliver is practical, responsible, deeply embedded in the workflows where people already work. That’s how I see the adoption happening and that’s how I see AI will drive real transformation going forward, especially when it comes to PLM.
Rahul Garg: I think those are great points, Shirish. I think that whole context of industrializing AI is how we are trying to adopt it. And the way I look at it is going into the future; companies are going to be looking at becoming AI first and everything that they will do. How can AI be used to do whatever the task they’re working on. That can drive a significant amount of efficiencies into their into their processes, into their work stream. I think it’s gonna be a very, very exciting future in terms of how people do the design engineering work as well.
Shirish More: Yeah. And then, you know, me being an engineer myself, you know, I can clearly see that AI is no longer an experiment. Means I would have not said this a couple of years back because, you know, a couple of years back people were questioning, oh, what’s this ChatGPT? But now it’s really happening. It’s becoming a trusted partner in the design and manufacturing processes. And our job going forward is now to shape it responsibly, responsibly so that it reflects the expertise, values, and creativity of the engineers it supports. OK, so that’s the plan.
Rahul Garg: Wonderful. Thank you so much, Shirish. It was wonderful talking to you on this very important and fast emerging topic and look forward to catching up on some more details with you on this.
Siemens Digital Industries Software helps organizations of all sizes digitally transform using software, hardware and services from the Siemens Xcelerator business platform. Siemens’ software and the comprehensive digital twin enable companies to optimize their design, engineering and manufacturing processes to turn today’s ideas into the sustainable products of the future. From chips to entire systems, from product to process, across all industries. Siemens Digital Industries Software – Accelerating transformation.
