Transforming gigafactory ramp-ups and reducing scrap rates with Siemens, Voltaiq and AWS
Show notes
We are wrapping up the first season of The Battery Podcast with a two-part discussion from the Battery Show and a couple high-profile partners from the industry. Puneet Sinha invites his counterparts from AWS and Voltaiq on stage to talk about the role of data-driven manufacturing in the battery industry.
Regulars to the series know that Puneet is the Global Head of Battery at Siemens Digital Industries, but his guests are Hari Gopalakrishna Head of Strategy, Sustainability, and EV at AWS and Tal Sholklapper, CEO of Voltaiq. Let’s dive into how data-driven manufacturing can bring long-term success to a sometimes-uncertain industry.
What you’ll learn:
[0:31] How three expert companies in their fields came together
[7:20] Why data-driven manufacturing is critical for the battery industry
[8:45] How a data driven approach can accelerate development timelines
[12:03] The business case for standardized factory architectures
Transcript
It's time for another episode of the battery podcast from Siemens Digital-Industries. We're wrapping up the season with a two-part discussion on data-driven manufacturing for the battery industry. We'll hear from our host, Puneet Sinha of Siemens, as well as experts from Voltaic and AWS to examine what we're working on to deliver more effective manufacturing processes across the industry. Siemens is partnering with AWS number one in cloud. And Voltai, which is the number one company when it comes to battery intelligence. So we are bringing number one company in industrial software, number one player in cloud, and number one player in battery intelligence, and we put our technologies and expertise together to achieve what I just showed you in the previous slide. Before I go and tell you more about this data-driven manufacturing, I want to emphasize that this is not a concept. We are delivering this to customers, and these are some of the benefits that we are. Helping our customers achieve 15% improvement in production efficiency because they have a digital twin of factory. They can understand how different machines are working, what are the problems they can solve. 2 times faster production scale up because they have the right ITOD convergence. And 10% or more reduction in scrap rate because they are able to leverage the data that is coming out from their factories and extract the battery-specific intelligence from that. To talk about some of these customer stories, what we are doing, I would like to bring My partners in crime, Hari Gopalakrishnan, who is worldwide head of strategy at AWS and Tal Shullapper, who is CEO and co-founder of Voltai. Welcome, guys. And what I'm hoping is rather than me talking, we talk through our collective vision and practice on data-driven manufacturing. How we are delivering this to customers and what kind of questions we are getting from them. But before we go ahead, I would like to go ahead and Hari, help us understand how AWS is looking at battery industry, what kind of values you guys are bringing to, uh, the industry. Hello. Um, first of all, thanks for having me. So, you've done a really good job of summarizing the challenges, so I wouldn't really dwell on too much. But I'll share with you some of the things that we're hearing from our customers. The first one is around high cost of experimentation in battery manufacturing. Developing new battery technology, new process. This is uh requiring, requiring a large amount of prototyping, lab testing, etc. um, and this is really expensive and time-consuming. The other thing that we hear a lot is around the scaling of lab to giga scale, right? Really Scaling up manufacturing while maintaining consistent quality. Across the production is a really complex challenge. And, you know, it's something that we have to solve. And the third one is around really long time to market. And high scrap rates. Of course this is, you know, you're producing millions of. Cells and you have to maintain a tight quality control across all of those. And uh that has to be done, but it's not a trivial task. So what you might ask is, hey, what is AWS's role in all of this? Uh, first of all, as you mentioned earlier, All of this generates massive amount of data, and this obviously on the floor, equipment, process, MES systems, uh ERP systems, PLM systems, they all generate data. But there's also data generated from materials, uh, from, uh, supply chain, right? So all of this data. You know, you have to make sense of this data, right? And so for us, AWS offers a suite of data services. Including advanced analytics. Machine learning, generative AI. And all of this to Make sense of this data, you know, generate insights from the data. And we have partnered with yourself, Siemens, and Voltaic. To bring in the domain expertise, so we consider us domain experts in data. Siemens brings expertise in industrial automation, connectivity, software. And you bring from V Volta battery intelligence, battery knowledge. And we want to bring all of this as an, in an, in an integrated set of technologies and tools to help address all of the challenges that I mentioned and help accelerate innovation and, and really shorten that time to market. Awesome. And, and we will, we'll, uh, delve into all of these things in a few minutes. But before that, tell, help us understand Voltaic and what you are doing and how you're looking at the battery industry. Yeah, so, a little bit of background, I started my career as a materials engineer designing what went into batteries and fuel cells and You know, from there, uh, you know, technology to partners who manufactured it out of the Department of Energy and then started my own company where we manufactured technology as well. And one of the challenges my team constantly had is trying to make decisions from all this valuable data on the production line, whether it was to identify a root cause, issues we start having in the field, where did those come from, um, and so we realized, you know, this opportunity here, especially as the battery industry is becoming. Uh, really as a critical foundational component, uh, in our modern, not just electronics, but also our vehicles and on the grid. Um, that we really needed this kind of technology to not only help accelerate the next generation of technology, but if you look at it today, you know, China has hit scale. They have hit economies of scale, their price points for cells, for vehicles, for packs, everything is, you know, dramatically lower because they have hit those inflection points. And for the rest of the economy here, you know, outside of China, the key is really going to be how do you get to those economies of. Scale, how do you get to those quality levels in order to be competitive in the market going forward and really one of the ways we saw to do that was to, you know, combine with a handful of key partners to deliver a really complete solution to the market that allows you to not only, you know, collect the data, you know, on the equipments and be able to bring that into a common platform but also start to iterate and learn a whole lot faster with these modern. Uh, you know, data infrastructures which involve the cloud and so bringing that together to help, you know, shift that left, identify problems faster, help identify root cause, help spin up these new factories in a much faster, more agile way so that you can hit those price points with these new products. Awesome, awesome. So, let's dive a little bit deeper. I talked about data-driven manufacturing, how we are looking at it, but I would love to hear your perspective on criticality of data-driven manufacturing for battery industry. So if you want to share something more, I'll start with you, Chao. Yeah, I was just on with the, you know, head of quality at a factory, uh, I got a call yesterday morning and You know, it's one of those things now as they're bringing on their next factories that things are pretty stable in the one they've been running for a number of years now, but with the new one, it's, it's all super manual. It's, it is a slow process, uh, you know, the reality right now is between when, you know, issues are identified and they're able to remediate them, you may end up having months that are spent really on the ground trying to piece that all together. And you can't take that long anymore. These multi-year ramp-ups, these high scrap rates or high return rates of stuff actually makes it into the field, uh, are just not gonna cut it as far if you want to sell at the low price points that are demanded by the market. And so, uh, data-driven manufacturing really gives you that infrastructure in place to, you know, not only collect that data, but actually learn from it and make it useful, uh, specifically around ramping up these factories and ensuring the quality of the cells and devices going forward. Awesome. Hari, how do you look at it? Yeah, great points. I'll just add on a couple of things. Uh, one, I would say. So why, why data-driven manufacturing, right? So you want to accelerate innovation in the space, of course, there is a significant amount of uh uh technology changes that are happening. Battery technologies are being developed, processing, uh, processing, uh, and manufacturing is also being developed at the same time. So you want to have the ability to rapidly innovate and to innovate faster, uh, you wanna have the data, right? You wanna have the data from, let's say manufacturing, supply chain, R&D, all of this data to be synthesized to be able to rapidly, you know, bring, uh, the product and also, uh, not just rapidly bringing the product but also making sure the quality to your point is, is, you know, where you need it to be. Second one I would point out is around process optimization, right? Um, you're collecting large amounts of data from across the equipment process, uh, MES systems, and how do you really, uh, take advantage of this data, really apply advanced analytics, machine learning, etc. and how could we generate insights that are actionable, right? I'll give you a good example. We have a customer that's in, uh, in the US who is currently using a classic, uh, vision-based system for material defect detection. And uh the system was generating a number of false positives and uh they decided to create a second AI based secondary detection system that um Required about 15,000 or so images to be, uh, you know, trained in, in the cloud, and the model was deployed at the edge, yeah, and that model, the challenge really wasn't about the model, you know, the model was doing really well, but on a fast line, the decision had to be made between 400 and 800 milliseconds. And so the secondary decision had to, to happen within that 4 to 400 to 800 milliseconds. And that's after the primary, right? And so both had to happen within that 800 millisecond time frame and so the result of this was that they had an improvement of 15% which for a factory producing about 5.5 million cells a day. Uh, you can imagine the ROI. So doing this continuous process optimization, it's a journey, and you cannot do it without data. Yeah, and I think one of the things to emphasize for those who are, you know, now entering the battery space, that this is a continuous evolution, you know, batteries are, uh, as one of our advisors described, who worked at Panasonic, Quantumscape, and others, leading manufacturing. It's, it's an unstable equilibrium. There's always something trying to push you off the edges, so it's not just getting that factory up and running, you know, day one, it's continuing to have that, you know, run effectively and it's new materials, it's new processes. We are, you know, continuously tweaking and improving things in these lines and making sure that runs stably is really, uh, key and that without data-driven tools, it really is a much slower and, uh, more costly process. Right, right. So, uh, at this point, I want to, for our audience's benefit, I want to show a simple schematic of how the three of us are coming together. It's a simple schematic, but, and I would love to hear your perspective, but let, let me take a crack at it. So at the very bottom, you see your actual factory, and your, your whole production line. As, as Harry and Tal also mentioned, if you're just doing things in a, relying on your machine builders or experience of your personnel, it is a very cumbersome process. What we are doing is we are putting a few layers of how you can build the right tech stack, uh, to do the things in a fashion and a scale that you want to do it. And it is, you can do it at the pilot scale, you can do it at a giga scale. So there are 3 key layers of it. One is creating a digital twin of your manufacturing that can be for the whole production line, can be for each of the machines, having the execution with MES, but also how you connect all the machines together. And with that, you are collecting the right data. And the data, then the middle layer becomes very, very critical. That is to have the right backbone of how you're collecting the data. And then once you have the data, you need to contextualize it. And the top layer is critical, but that only works when you have the bottom two layers, and that's what we call battery-specific intelligence based on the work from Siemens, Voltaic has enterprise battery intelligence. AWS brings those AI services. So you make decisions. First, you understand the data, what the data is telling about the things that are happening at a microscopic level inside a cell, and then make the right decisions of how you need to go and tweak machines, operations or something about material and feed that information back. So AWS also provides that backbone to make all of this happen. Anything more you guys want to add to this? I'll give you an example to make this, you know, real. We're doing, uh, uh, we actually have in production, implemented a solution for a customer that has, uh. Needs around training that was taking almost about 9 months to train new employees. Uh, what we did was train, uh, essentially bring. All of the data from all of the different systems of records like MES and uh systems like for example, like your uh standard operating procedures, uh, your data from your operating, you know, your, your, your manufacturing floor, uh PLM systems, logs, uh, data from, for example, sound data, video data, so structured and unstructured data, bring all of this in to train an LLM. And that LLM was essentially cutting, cutting down that, that training time from 9 months to just. You know, just a few months, uh, like significantly cutting down that time. So this, you know, especially in an industry that is struggling with, uh, you know, labor shortage and technical skills shortage, these kinds of solutions can be, you know, can be the, the top, you know, that takes advantage of all of these different layers and all of the different functionality that, that you're seeking in different, different levels to come to life at the top. No, and If you look at this, you know, ecosystem, it, it sort of makes sense, but it's one of those things that's trivialized as you go to a lot of the new factories as they're being rolled out. It's, you know, we haven't done this before. We've got to get the grit, the land, we've got to get the equipment. This is sort of an afterthought. We'll like put it all together later and we're seeing the repercussions of that at a lot of the factories and facilities that are out there that are struggling to scale now and It is not trivial to, you know, network your equipment to get the skate up and running, to get the MES to get the data into analytics infrastructure so you can learn from it and really what we're trying to do here is provide, you know, really a reference architecture. There's always gonna be a little bit of customization, configuration, different tools you bring in. Uh, but really have this reference architecture that we can take to customers and help you get up and running very quickly rather than have to take years, uh, struggling to piece these things together as you've already hired 1000 engineers, you're putting material through the equipment. And, you know, one of our customers talked through, it's a relatively small factory and for each year of these delays they're having, you know, it's $250 million. It's even for a tiny facility and so the, the opportunity here to have this packaged up and ready to go as you're starting to produce is really important. This seems like a good spot to pause the conversation and take a break for a little bit. Thank you, Tal, Hari, and Punit, and thanks to the audience for tuning in to this episode of the Battery podcast. We'll be back soon to talk about the structure and business architectures that make data-driven manufacturing possible for the battery industry. Make sure to subscribe so you don't miss future episodes, and until then, be sure to check out our website Siemens.com/battery for more information across the entire industry. This podcast contains the personal views and opinions of the speakers. It is important to note that these opinions are solely those of the speakers and do not reflect the views or positions of Siemens, its management, its employees, or any entities the speakers may be affiliated with. Siemens as a company does not endorse any of these statements made in the podcast. Siemens bears no responsibility for ensuring the content's accuracy or verifying the statements made in this podcast.