Accelerating industrial transformation with IoT on AWS

with my counterpart brad davis from toyota north america. uh we're gonna walk through lots of content, tons of things to tell you about today. um but uh i thought i'd throw together a little agenda.

uh first off, we're gonna talk a little bit about why digital transformation is still so challenging. and the honest truth is it is really difficult. still, even after all these years after that, i'm gonna share some lessons uh that we've learned working with lots of customers who are having big success in their digital transformations. so hopefully, we can kind of take what we learn from them and share it with all of you.

um and of course, it wouldn't be reinvented without some launches. so let's have some launches. we'll do some launches there and then uh we'll hear from toyota. uh my, my counterpart here, brad will come up and talk a little bit about what they've done at toyota north america.

and uh i've also got another customer called nsg who's had some really great success lately, uh really cool system. they were willing to share with us. uh we've got a little demo of that and uh i'll share with you what nsg is doing.

and if you all stay to the end, um we have uh an exclusive uh demo implementation of uh a gen a i uh shop floor assistant. and that's really cool. uh this is sort of the, the leading edge bleeding edge. and uh we wanted to share something with you that is immediately accessible if you're interested in recreating something like that for yourselves. this is all open source and we'll be able to give you a link to that at the end of the presentation.

so if i, if i have a hook to keep you to that last 59th minute, that's it. so, like i said, digital transformation, i mean, we've, we've been talking about this for years. there was the uh advent of industry 4.0 that was back in what the nineties that came out, uh early two thousands. we really got serious about it later in the 20 tens.

we, we had a lot of uh sort of resurgence new hope because iot was gonna be the next big thing that really unlocked uh uh the industry for auto promise of being anywhere uh at your child's swim meet and use your phone to completely, you know, uh a act an alarm, uh dispatch a crew and, and fix aaa running uh problem in, in your plant. um but truly, it, it's still very difficult. uh i mean, the, the benefits of, of digital transformation are absolutely huge and our customers are seeing big results.

um there's a lot of examples out there of people getting, you know, huge gains on uh worker productivity. we've got a great example that i shared last year with koch industries who's doing uh a fully connected worker program, digital twin. uh they're rolling that out to multiple sites now. um really great uh improvement in their worker productivity there.

uh we've got, you know, uh throughput increases, uh reduction of downtime that are are significant, really statistically significant and have a big effect on the co quality and the cost bottom line. and of course, then the co cost of quality improvement as well. we see a lot of really great uh progress towards customers, you know, meeting those types of business goals with digital transformation.

but the truth is we still see more than two thirds of these transformation initiatives stall in just the pilot phase. and why does that happen? uh we have a number of theories and today i'm going to share with you a lot of the conclusions out of those theories. but i think first and foremost, a lot of these digital transformations efforts fail because they're not aligned from shop floor to top floor. the itot gap is still very real.

um and often these things start by somebody in the c suite getting uh pressure from a board saying, hey, you need to modernize, you need to be digital, you need to have a digital transformation program. uh or we see some really great pilots that start out of, you know, innovative people working on the shop floor doing continuous improvement and gathering data and, and thinking big thoughts but not having an outlet, not having a way to take that forward beyond their work center or their area or their site. right.

so a lot of these things kind of get stuck, they also get stuck because of things like we didn't think about security upfront. we didn't necessarily align it to our actual operational business goals, even though we thought the technology was super cool. we didn't, you know, feed this back into our lien management system, for example. so a lot of these things can really just die on the vine in that pilot phase and they're not moving forward.

but what does it mean to be truly digital? a lot of these things, you know that again, they come out of out of multiple different aspects, sometimes they're driven by it, sometimes they're driven by the, the ot the shop floor people.

um you know, we've got sustainability goals out there pushing us to think about how are we capturing our data? how do we know, you know, how much carbon are we producing per ton of product that we're also producing?

um we've got security concerns. i talked to somebody this morning who said, yeah, i'm, i'm looking at doing, you know, shop for digitization, but i still, i'm not sure that i'm ready to go to cloud yet. right. so there's, there's still some hesitation there and, and you know, in exploring why there's a lot of problems like, you know, the, the introduction of it,

um this separation between the shop floor and the it folks who need to think about standardized things across their entire operation versus the people on the shop floor who are saying, hey, is this going to impact my job? how is that going to impact my job? is it gonna make my job better or worse things like that?

but ultimately, taking a data driven approach and really thinking through your data maturity and thinking about how do i get all the data, how do i contextualize all that data? how do i use that to make real business improvements? that's when we start down the proper digital uh journey.

and i think my definition personally is, you know, to be digital in this space, you truly have to bridge that gap between it and ot to gain a really unfair advantage over other people who are not doing the same, right? and this might come from a productivity gain, a performance gain, sustainability gain, an efficiency gain.

this is gonna happen and, and we need to drive this forward, but it is gonna start by bridging that gap between it and ot so what have we learned well back when we started all this journey, you know, we were kind of in traditional manufacturing, low automation, uh low data uh visibility and we started censoring things. we started adding as many, you know, vibration sensors and heat sensors and, and trying to get access to the data of old machines, brown field things using things like mod bus and, and other protocols. but it's, it's really challenging at that point.

and again, this is why a lot of these things fail. it's hard to get machines connected. but as we've moved through, we've, we've kind of moved to this spot where i i consider this is my own chart, by the way, my own view of the world, but we've moved to connected manufacturing. i'm not sure we're fully at smart manufacturing yet, but we are at least in connected manufacturing.

that means we've got a reasonable level of automation. we've got a reasonable level of access to data and we're starting to see the benefits of combining these really thinking harder about how to make improvements, you know, looking for our, our biggest uh downtime things like that just just quantifying little things like that to align better to business goals.

but the promise of this could be much greater if we improve on the automation and we improve on our data maturity. and really think about how to get that, that granular data, the contextual data that puts all of this into focus to tell us where our real problems are we can move very quickly to something like software defined factory or software driven automation, things like that.

and who knows? i hear a lot in the industry uh in the, in the press about lights out manufacturing, whether you believe in that or not yet, um we could be trending there as well. but if that's at least the ultimate goal, lights out manufacturing, how do we get there in a clean way?

so aws has, has built uh a portfolio as you know, if you've seen me present in the past or if you've uh been to reinvented in the past, you know, that we have uh a great suite of really foundational services, things like iot core, iot green grass, uh kinesis, video streaming. uh we've got amazon sage maker, uh manage grafana to make nice dashboards.

um uh iot device management, things like that. these things are general purpose and great for connecting uh you know, equipment or machines, things like that. but when we moved to, you know, really look at the customer problem and we started working backwards from those customer problems. we figured out that these services are foundational and we, we continue to need those in our architectures.

but we also needed to augment that with some of the things that are more like our experience prebuilt uh patterns that we saw happening over and over. the customers were having to solve themselves, things like hot, cold and warm tears, data storage. in the cloud.

um you know, common patterns for managing time series data, log file data, uh instant alarm data, event data, all those types of things that are not that common to a traditional iot use case but are super common to industrial cases and it really doesn't matter what industry you come from.

um we're really blessed that we get to sit right at the nexus of people from food and beverage, oil and gas, mining, uh manufacturing, automotive, um all sorts of different uh uh walks of life, all sorts of different parts of the economy, all having the same problems when it comes to managing data gathering, data, connecting the sources, making sure they can act, act alarms, uh helping them to really quickly get to issue resolution when there's a problem and therefore, you know, increasing up time, decreasing downtime.

so we've, we've built these purpose built services that take a lot of those common patterns that we saw into account. but where we've really hit our stride, i think you've probably heard about this stuff in the past where we've hit our stride is when we realized what our purpose was, we can't solve all of the problems out there.

we're just not equipped, we're not in the machines, we're not in the pl cs, we're not in the drives. um we don't have deep process expertise on, you know, how do i optimize a, a dairy pasteurization uh or a clean in place process.

um we just don't have that expertise. what we do have is expertise on managing data in the cloud at hyperscale for multisite lots of connectivity, high volume asset intensive data. and when we brought all of this together, we started thinking harder about how do we really provide value to our customers.

and it's really looking at these prebuilt architectures, not just prebuilt services or not just prebuilt function blocks but prebuilt architectures for doing things like industrial data management in the cloud. and this is why our second lesson that we've learned that i can share with you digital transformation is absolutely a team sport. nobody can do this themselves.

um it's, it's very rare that we find a company who has every single aspect and is able to do that. and if they do, it's very rare that we find a customer who says, yes, i'm going to lock in end to end with one single s single cus uh provider, uh one single vendor, it creates risk, there's other problems.

so really, we've taken this approach of saying, hey, aws needs to act as a team. we need to work with a great partner network. we can build extensible applications together. we can think about how to provide these industry patterns, these common pre-built patterns that we can provide, that we're kind of uniquely positioned to provide with our partners.

and we can focus on things like cloud data management for industrial type types of data uh as our kind of core value. so we've worked with partners in every sector, every walk of life, every aspect of, of, you know, from the silicon to the connection, we call it the first inch problem getting data out of devices.

uh you know, there's all these uh what tens of thousands of industrial protocols from the last 20 years, we have equipment that was commissioned uh from site to site in different decades, depending on uh you know, perhaps when it was acquired, perhaps when it was built. and o often these systems all have different uh uh original vendors, right?

so we work with partners who know how to deal with that, who know how to connect to those things and who can, who can really think through uh where their value comes from companies like k uh i know we've got uh taro in the audience, k is building uh solutions for moving people. uh they're not just an elevator company, they're moving people, they're, they're thinking big thoughts about how do we make office buildings more efficient? how do we make spaces more efficient?

um so con is a great example of this. uh you've heard us talk about siemens. um you've heard us talk about accenture. um what you might not have heard about is, is some of our other partners, some of our smaller partners who do huge things, a company like embassy of things that we're working with at hill corp h corp is a, a company that buys up wells well sites that are late in their life cycle.

and so, uh they're trying to maximize the remaining production out of these old well sites, as you can imagine as a company model, this is an it nightmare and an ot nightmare. um all of these well sites have come from different companies. they often have different ska versions, even if it's the same company, they have multiple different historians, multiple different ska systems and really no single source of the truth.

so when hill corp came with us and embassy of things, their whole plan, their whole need was to get one source of truth, multisite, quick visibility across their entire operation across all these disparate sk systems. all these different historians and embassy of things was able to help us with that solution.

um they rapidly acquired uh pardon me, they rapidly assimilated all these assets. um they created a a uni unified visualization across the entire operation. and by doing all of this, they actually managed to reduce their it spend by a significant amount and increase their production because they had better visibility into what their wells were doing.

we also worked with a company called reply storm for our customer, kingspan. um kingspan is a, a big uh building company. they've got about 50 factories worldwide and when they looked across their entire operation, it's, it's the classic problem that i described multiple different systems depending on when they were acquired, when they were built, when they were commissioned, uh and across 50 different factories. you can imagine when they tried to do an audit of where are we at? what does our supply chain look like? things like this? it took them three weeks to do reply storm helped them build a much more data driven uh application with a really great data foundation that was repeatable and common that allowed them to cut their audit time from three weeks down to three minutes, which is really impressive.

um it created data driven decisions that were not just at the shop floor but all the way up to the board. and uh reply storm also helped them to make this very, very repeatable. so they had this iot in a box solution that they can now deploy at new greenfield factories that they bring online as well.

so great working with these partners. and i'm really excited today to talk about uh one of our, one of our partners that we've been working with over the last couple of months to make things really accessible, really easy

Uh I talked about that first inch problem. It's very difficult, often dealing with so many different types of industrial protocol. And uh that first inch problem getting the data off the device is something that is, is one of the main blockers of moving digital transformation forward.

So we've partnered up with Easy Edge uh by Dematic uh to bring 10 newly supported protocols into our Sitewise Edge uh offering. And this means now we can support Modbus, MQTT, KNX, uh the Siemens S7, Prophy Bus, Prophy Net, Ethernet IP, uh Laura WAN, and even RESTful APIs. So this is pretty cool.

Um this is now available today. We launched this on Sunday and uh it allows us to, to more quickly connect to all those brownfield machines. One of our very first customers who used this was Siemens Energy. Um this is his quote. Uh he said "It's a game changer to connect more brownfield assets in less time and at a lower cost." So we're really excited to uh talk a little bit about Dematic and our and our machine connectivity framework.

Another relationship we would have heard about lost. I'm sure in the last couple of years we've had Siemens on our keynote stage talking with our CEO Adam. Uh and so, you know that we have this, this uh very fruitful and, and very uh good relationship with Siemens. Uh and this is our, our premier partnership really for that uh really end to end working with an automation provider who's also able to provide the applications, the industry knowledge, the deep expertise that, that we need when we're doing this as a team sport.

Um back at Hanover, if you weren't at Hanover, you might have missed that we actually launched uh a brand new solution called Lean Daily Management. And this is with Mendix. Um a combination of Mendix with our digital twin product called IoT TwinMaker. Uh Lean Daily Management allows uh people on the shop floor to really quickly connect across multiple disparate systems. Get that visualization into a twin, be able to create items and visualizations and do it all with the Mendix low code environment so they can quickly access their data.

Um this allows them to, to cut that data acquisition time from again, weeks to a couple of hours. Uh and uh uh really provide that contextual data uh up to that low code environment. Um this is a game changer for the shop floor so that we can really bring that visibility quickly into the, into the shop floor. So uh we're really excited back in Hanover to, to launch this Lean Daily Management suite. I would really encourage you to go check it out. There's a QR code on the s on the side there. Uh and it brings together both the the twin aspects along with that low code aspect.

Something else I'm really excited today to talk about. Uh I think we launched this one on Sunday or Monday as well. Uh but the uh Sitewise Edge from AWS is now available in preview on the Siemens Industrial Edge marketplace. This is a really big deal, this one for too long. It's it's been very difficult to bridge that ITOT gap and to make sure that we can get data quickly out of machines and really move data from that machine edge up to the cloud and get, you know access to things like machine learning, think about, you know, multisite auditability, traceability, sustainability programs, things like that.

And often really Siemens being the leader on that industrial shop floor, uh the machine edge and and building the edge. Um you know, this is a great solution but it's really been geared towards OT people. Uh we have this great solution called Sitewise that allows you to contextualize data, analyze data and really dig into your, your cause codes, things like this. And to bring these two together inside that same framework of this Siemens Industrial Edge uh is, is really game changing.

This allows us to connect those machines, those devices, all that data super quick log that right up to Sitewise. Um this is available with a cross uh cross marketplace interaction and uh to learn more. I really encourage you to stop by this uh this uh joint breakout session tomorrow. Uh it's at Mandalay Bay. It's at 8:30 a.m. but uh we'll have Torben from Siemens and Nicola from AWS, talk more about this solution and the power of having these two things working together. So, really excited about this one today.

And so Siemens and, and AWS, we continue to try to innovate and we continue to look for things like how can we use these industrial applications that were built, purpose built for manufacturing for automotive, for dairy, for food and beverage, things like this. Siemens has that as part of their accelerator portfolio. Um Mendix is a great tool that's now available. Uh also in the siemens industrial edge working again with Sitewise, we've got a, a Mendix to Sitewise connector as well. So you can now use Mendix, Siemens Mendix in conjunction with IoT Sitewise very easily, just, just download the connector.

But we continue to innovate together. And what I really love about this is this is an end to end reference architecture of the two companies coming together. We've got the leader on uh industrial automation and uh uh uh sorry, I thought there was a hand up there. Uh the leader in industrial automation uh working with uh us in a leader in the cloud, uh working on that hyperscale cloud management of your data uh and really bringing this end to end architecture together and we're gonna continue to innovate for, for our customers and, and uh work together in this space because again, IoT is a team sport.

So digital transformation, this one's really obvious, but it's uh greatly accelerated when we have all the data. I know that sounds really silly. Uh of course, that would be true, but it's so true and it's so important um to start looking at, you know, how do we gather that data? Where is that data coming from? Do we have enough of the context data. Are we passing that up to the cloud things like this to do the the deeper analytics?

And one company that was working on this was Bristol Myers Squibb. Um they had uh an objective to uh accelerate their business strategy, uh looking enterprise wide and really start to combine these systems. They had some of their data trapped in different PI systems. Uh I'm sure very useful to the site PI is a great system but uh at the same time, not able to give them enterprise wide visibility.

So Bristol Meyer Squibb was looking for, you know, end to end traceability and a single source of the truth. Um we worked a lot with Bristol Myers Squibb and other customers in the last year. And we've really thought about, you know, how do we get this data more accessible, more contextualized and more quickly up to the cloud?

Um and so today, I'm, I'm also excited to talk a little bit about some of the enhancements that we've done for IoT Sitewise uh to help that. Um when we think about Sitewise being this cloud data management uh service for industrial data, you know, we really thought hard about how do we represent data in the best possible way. And Sitewise uses models, these equipment models, asset models to represent that data, make sure that it's consistent across system, things like that.

But it was still very difficult, very, very manual uh to update these models to create the model uh and things like that. So we, we thought deep thoughts about, you know, these, these small increment uh incremental innovations that we can bring to our customers, but that make a big difference.

So today, it's exciting to tell you that we can now import asset models and hierarchies from existing systems, things like, you know, historians skas, things like that and we can represent those now inside of Sitewise with very minimal uh amount of code. Um so we can import those models and use those.

We've also created this uh concept of reusable components so that when you're trying to derive a new asset model or launch this at another site, another facility, um we can, we can do this in a very, very quick way. And uh uh ee even in, you know, taking into account the variations that might exist site to site, we've also really spent a lot of our year thinking deep thoughts about one of our main customer pain points that we've heard a lot, which is, you know, I don't want to pay the cloud tax.

I'm already paying all this infrastructure and I'm building all this stuff locally and I've got a lot of data on prem and when I send all that data real time to the cloud, it costs a lot of money. So we've spent a lot of this year really thinking about, all right, how do we move data, how do we move those bits more effectively, more cost effectively?

Um thinking about different patterns and, and uh and different things that we could do. So, I'm, I'm happy to tell you that we've thought about this. We've, we've launched something called adaptive ingestion. Uh with that we believe will reduce the cost of moving data to the cloud by about 70% and that's tested.

So, um we've thought about this in terms of creating a warm tier where data is immediately accessible. Uh we still have our, our uh low latency uh instant access time series uh streaming as well for any of those. And we still of course, have real time events, real-time alarms, things like that, but a lot of your data doesn't need to move in real time.

A lot of that data is there for audit purposes. It might be there for uh you know, daily production max mins. Um it might also be there as as 15 minute interval data to help with your sustainability program. Um something like that. So when we think about 15 minute interval data, we can send that data to a warm store.

We can do that 96 times a day instead of thousands of times a day, we can do all these roll ups in the edge. Thanks to Sitewise Edge, we can move all that data in a very, very optimized fashion to reduce the cost overall. And as I said, our results are around 70% which is great.

And we've also thought really uh deep thoughts uh again about now, now that I can get data up there quicker and I can do it much more cost effectively. How do I now get it out just as easily? And we, we realized we had a problem that uh it was actually kind of difficult to get data back out of Sitewise if you weren't using one of our own tools.

And so, because we kind of recognized, hey, this is a team sport and we want to make sure we're working with the best in class application providers and ISVs that are out there. Um we thought harder about, you know, making sure that this is available through simple query interfaces. So we now have a SQL-like interface that you can use to pull data from Sitewise. Seems like a simple thing. We probably already should have had that. Uh but we have it now. So very excited to tell you about those.

And this year we've, we've launched around 15 new enhancements to Sitewise overall. So it's been a big year, all of them incremental, all very small, but they have a big impact overall.

And for Bristol Myers Squibb, uh you know, this is the quote uh from, from uh Bristol Myers themselves. Um but they've been able to uh extend the exposure and the consumption of the data scale beyond the personnel and the point to point solutions, they're getting true enterprise wide visibility. Uh and the long term vision is now to enable additional use cases like predictive maintenance and, and things like that.

So they're now looking at using this data along with some of our AI and ML models to do predictive maintenance and, and uh and get more value out of their data.

So I would like to now invite Brad a real life person who has done this digital transformation at Toyota. And while Brad's coming on stage, we've got a short video from uh from Toyota on the field.

Thanks, Brad.

Hi. My name is Steven Palmer. My name is, my name is I'm Corey. I'm a skilled team member here at TMMK. I'm a maintenance team leader at one main. I'm a team member at M. My favorite is always wear a seat belt. Well, let me thank you

always ensure that my team members are performing their job safety and my safety commitment is to always follow, walk outside. my safety commitment is to always be safe by choice, not by chance.

so this new system has really improved our troubleshooting process. um just by the simple fact that we can look at the, the data, the raw data coming from the machines and actively see where issues are coming. it's easier for me to find the root cause of the problem. instead of looking at the symptoms, you can pull up multiple source of data from a machine to see what the problem is.

it allows me to use proactive maintenance with real time data and predictive anomalies. so i can schedule work to be done during downtime. also with limited parts and lead time to get us parts, we can adequately use parts on a machine that we know is selling than to use on a machine that we think is failing from that data. we can see what the issue is and go and look at it right away before it breaks where production works three shifts during the week.

this system allows him to troubleshoot without causing any downtime and allows me to plan work for the weekend accordingly. so our focus line is t three c a and the o a on t three c a since we've been doing this for the last month is, has went up from the high seventies, low eighties to low nineties highs. it's, it's really increased o a on t three cm.

so this system really does help schedule plan work when we can go in here and find severe alarms on these machines. we can that, that gets us proactive and going down to spare parts to confirm. we have parts for these machines the way we can go ahead and schedule for weekend work. if we have all the parts, we have everything ready. we've got guys signed up for the weekend. we're gonna go in that machine. we're gonna investigate it. all right.

so thank you, mike. i am brad davis. i'm a engineering maintenance manager in toyota motor manufacturing kentucky. uh specifically i work in the powertrain plant. so we build engines for toyota camrys, rav fours and lexus ess. so the people you just saw in that video, those are all my maintenance team members. um and they were using a system that we built called reliability insight.

so, what i'm here to talk about today is our journey with amazon web services and how they helped get us to where we can have that reliability insight system for our predictive maintenance.

toyota as a company has been working for a long time in industry 4.0 optimization trying to develop new systems within the space, especially io worlds uh within manufacturing and especially at tmmk powertrain. our focus is on machine reliability.

so back in 2020 with the maintenance organization's goal to create systems for machine reliability, we started the maintenance 4.0 working group. so we created the vision of a hot topic, right? data driven decisions. but if you look at those three words, they're all three very important words.

so data is we wanted to utilize our existing systems. so like many companies, toyota has years of new years of data collection systems for different types of data, for different types of equipment. and it all goes into servers and databases but is never truly optimized for predictive maintenance. so we wanted to leverage that data.

and then for the industry 4.0 systems, next thing is driven. so our main goal is to empower our team members to make uh informed proactive decisions, to do the right maintenance on our equipment at the right frequency. and the last thing is decision. so leveraging the machine data and that driven mindset from our team members to increase productivity.

so to increase productivity, our main goal was solid predictive maintenance systems and predictive maintenance is important.

so where we started our journey is that the vast majority of equipment downtime in toyota comes from unplanned machine breakdowns. uh that's our unplanned corrective work orders up there where we're going through a solid predictive maintenance system is that we can do the correct planned maintenance on non production time to prevent downtime during production, increasing ro a and reducing downtime.

so this is where aws came in. so we had a lot of data, we collect uh multiple monitoring parameters worth of data from our cnc equipment. so in power train, in the machining line, the area that you saw those team members. uh in the video, we have hundreds of cnc machining centers, they mill grind drill, uh all of the heads blocks, camshafts, crank shafts and connecting rods for the engines that we build.

so each piece of this equipment collects data from our pe cnc controllers as well as our pl cs from there that date goes to an edge gateway that converts that to the mq tt or o pc u a protocols where it's sent to uh green grass and then green grass sends it to sitewise.

so in our architecture, sitewise acts as the um data historian as well as a unified name space for all the other aws services within this process. but that sitewise data then goes to uh look out for equipment. so it gets there through sage maker pipelines pulling from athena data tables built on the um input data as three buckets. but l four e gets input data and output data.

after the lookout for equipment processes, that data sends it back through uh that sends the inference results back through lambda functions into sitewise. so you see maximo, ibm maximo up there as well. so ibm maximo is our enterprise asset management system. that's where we have all of our um maintenance work order logs. so it's our cm ms and using being able to integrate that as input data into l four e along with our sitewise data, we're able to generate the uh reliability insights system that you saw in the video.

so i'm gonna go into a few of the dashboards within that system kind of explain how our team members are actually able to use that to make their data driven decisions.

so the first is an area level view. so we have um multiple lines in each area. you can see that the parts that i talked about up there, your blocks, your heads, your cams. so this is a quick overall view of an area. it's used by leadership mainly. uh it shows the current operation availability, your overall equipment excellence, as well as aggregated downtime and the last column, which is the coolest is how many abnormalities are detected within that line.

so as you drill down into the line level, you get to see all of your assets with a few extra kpis as far as cycle time and some uh meantime, to repair and meantime between failures, some maintenance, kpis that help uh measure the results of your predictive maintenance systems. but um these are all sorted based on if there is a an anomaly detected on the equipment and how severe that anomaly is.

and this is where our team members can click on the severe anomalies and start looking deeper into what that means for that piece of equipment.

so this is a list of monitoring parameters. it's kind of a standard cnc machine layout. it includes uh servo axis, temperatures, insulation, resistance, average torques max torque, um servo health data. ultimately, that's coming from our pic cnc and team members can see which of these uh see the trend line of each of these parameters. and then they can um select those and see a graphical trend over time, compare that to other machines and use that as a troubleshooting tool.

the next portion in this asset level view is uh the integration with ibm maximo. so it is unplanned work orders. you know what has recently happened on the equipment, these unplanned work orders work as labels in l four e. so you see your equipment, health index and your performance. but l four e recognizes that when we have a work order here, the current condition of all those monitoring parameters is indicative of unhealthy piece of equipment. so we've got a problem here.

so when it sees that same pattern in all those modern parameters in the future, that's an abnormality and that'll show up in the predictive insights. so that integration is great here as well because as team members use these, this data, these trending, these predictive insights, they're able to create a work order straight from this screen and uh schedule that predictive maintenance on scheduled equipment downtime.

so what all this means for tmmk powertrain, the good stuff that's come from this.

so in our focus area, we've been able to prevent over $80,000 worth of unplanned downtime and spare parts cost being able to uh transition from unplanned to planned corrective repairs based on the predictive insights that we've gotten from l four e. in that focus line, the very first month that we started monitoring with this system, we were able to increase our operation ava availability 10% compared to the previous 12 month average.

um that fluctuates up and down of course, because this is not, the predictive maintenance is not the only solution, but it is uh we continue to see an increased trend in our operation availability uh based on the usage of our reliability insights. and then within tmmk powertrain, we've scaled this system to over 300 assets but even bigger.

so toyota motor, north america has selected reliability insights as the regional solution for a predictive maintenance system um in the industry 4.0 space. so we're currently working with plants within west virginia, alabama, tennessee, our vehicle plan in kentucky and texas. uh there's many working groups working together to scale this same solution across the north america.

let's give that back to mike.

thank you guys. thanks. yeah. so i wanna thank brad and uh also his counterpart braden in the audience for, for creating this system and, and sharing it with us cause uh we don't, we don't often get that. uh there's a lot of trade secrets out there. there's a lot of sensitivity to data, there's a lot of process that that is protected. so thank you for, for sharing that we really appreciate it.

um so i've got uh just two more great examples for you. um it looks like we're right on time. so, um you know, digital transformation, i, i just talked about this at the beginning. a lot of them, you know, two thirds stall in that first pilot phase. and what we find is that a lot of the time if we go on and, and we talk to people and we say, all right. so, um what went wrong? what, you know, it's kind of a postmortem. uh what do you think happened? and often the, the data that we gather at least says that people on the shop floor didn't necessarily understand the goals that were set at the sea level, uh or vice versa.

um you know, the sea level didn't necessarily understand something like what's coming out of kentucky at toyota. uh enough to say, wow, we should adopt that regionally, right? um so it's a, it's a big deal to, to really think through shop floor to top floor alignment. and really what we find is that as soon as you align these business goals to the digital transformation, everybody kind of knows their job. and if you can put this into language that the shop lawyer understands root cause analysis, lien management, things like that, um you get a much, much higher success rate out of these digital transformation projects.

um a great example here is nsgnsg is uh nippon glass. um we've been working with their industry for do o champion for a number of months. uh uh named peter lai peter uh wanted to be here. uh unfortunately, he couldn't. so he's recorded a great video that will make available online and i'll show you some of his screens as well

Um but nsg was really they're a global leader in uh glass production, especially glass cord and they were having a lot of troubles with uh breakage, uh a lot of product quality issues. Um they, they had a lot of downtime that were kind of unplanned, unexpected and they were trying to quantify and figure out what is, what is happening.

And so of course, they turned to peter to say, hey you with all the data and all the fancy tools uh tell us what's going on here. And so peter uh created a unique solution here uh with it. Well, in fact, it's become a very common solution. But uh with aws iot sitewise, uh working with amazon, look out for equipment. Uh this is what brad kept referring to as l four e uh and hosted into these managed grafana dashboards.

As you can see these dashboards are very complex. Um but what i really loved about this was, you know, the visibility that they managed to gain and peter was able to give their uh board level and their sea levels uh direct uh uh data um that everybody was looking at as one source of the truth.

And my favorite screen is this one. This is where peter, you know, he talks a lot about, we were trying to figure out where our quality problems were coming from, where our downtime were coming from. So they managed to look at every root cause and create a pareto chart out of that. And they, they started doing that so that they could say, all right, where are the biggest things that we should go investigate, you know, stuff like that.

But then they realized they needed more data and they needed better metrics and they needed to understand exactly what was going on behind the scenes. So peter uh introduced this uh uh look out for in lookout for equipment integration and very quickly you can see here, this is kind of their first oven uh o and a has very few anomalies. Um as, as he looked at oven b, tons of anomalies. Uh what this screen is showing you is, is a complete difference between that oven a and oven b uh and looking at where these were coming from, they needed to understand exactly what was happening uh through the entire process.

What they figured out was uh if i can jump to there we go, they built a digital twin and they started to monitor more tags in real time and really look through their entire process. And what you're seeing here is a combination once again of us uh being in a team sport, this is matter report on the screen uh giving a real life, you know, human eye relatable visualization. Um not just a, you know, cartoonish digital twin like i'll show next, but using matter report, they can actually look and say this is the exact place that this anomaly is coming from.

And peter was able to really quickly whenever they had a kind of a uh a thought about, hey, where's this uh this issue coming from? He was able to very quickly add an additional point through twinmaker to uh of data uh coming from sitewise, adding that point to twinmaker and then running that through this l three e integration um to find out where anomalies were happening.

Um peter's conclusion out of this, by the way, uh the one oven was acting completely uh abnormal. Uh it was also burning a lot more gas than they predicted. And when it was benchmarked against its pure oven or ovens of a of a similar size, uh the, the consumption was significant. So just by looking at all these root causes and starting to think through their data and really getting the context of all that data in the plant, they managed to save more than 100,000 uh pounds uh pound sterling. They're a uk company um £100,000 in just 12 weeks by, by correcting that gas overconsumption.

So something else i'm really excited to talk about today. Um we've made this generally available um but we've created a much tighter integration between iot sitewise and lookout for equipment. You heard uh brad talking about this lookout for equipment and sitewise integration uh and some of our other customers like nsg, but unfortunately, for them, uh it was a very difficult process. Um they had to extract data, there was a lot of coding to be done. They had to move data around. And you know, honestly, these two services just weren't really built to talk to each other that well.

Um so we, we spent a lot of time on this, this year. Uh we went back and we looked at ok, this is a common use case. Our customers love this. Let's make sure this works really seamlessly. And so we've created this really nice tight integration between sitewise and look out for equipment. You don't have to move data around anymore. Uh you can use sitewise to train these models. Uh you don't have to do a bunch of code integrations, anything like that. You can just build models for the anomaly detection uh and get that predictive maintenance data back into iot sitewise.

So, really excited about this one again, an incremental innovation this year. Uh we've got a lot of these little, little things that we've done just to make the systems that much better. But uh uh we're really, really happy with this one, check out the session thursday at 2:30 p.m. Uh iot 203 to learn more about this uh and, and see it for yourselves. But again, you know, brad's demonstration plus uh peter's demonstration, i think it, it shows the power of these things. And, you know, uh this is where, uh you know, peter says, you know, they've saved in excess of £100,000 in a remarkably short time frame. He tells me it was about 12 weeks uh by, by using these systems to benchmark that peer to peer.

So obviously, working in conjunction with our partners, uh and in the process and making sure we gather all that data, making sure that we have all the data that we need, you know, treating iot like a dreams, uh permit uh digital transformation and iot like a team sport and really thinking about do we have enough data? All of these things have really come together in these use cases to help these customers? But don't forget overall, the digital transformation, in spite of its name is not about technology.

Um it doesn't matter what great technologies we provide unless the people on the floor are actually going to use it and unless it actually starts solving business goals and unless it actually drives performance improvement. So this is a very human centric um uh transformation project.

And uh something that we thought deep thoughts about our, our customers were coming to us saying, hey, mike, we're hearing about gen a i, we don't know how to use it in production. Um we're hearing that this could, could be a game changer. Uh have you thought about where you could use this things like this?

So we did our own surveys and we really thought about their customer, our our customers and where the customers were gonna benefit the most from using something like uh generative a i. And we came up uh in just a very few short weeks. Uh one of my very talented engineers, uh principal engineers, giai uh created this uh in just a very few short weeks and this is now available to you. I'll have a qr code at the end as an open source project. But what we thought of was could we use generative a i to quickly diagnose an issue?

For example, here, our cookie factory demo which will be available to you. Uh we have a misshapen cookie. So we can say, hey, what are some of the common problems that would cause a misshapen cookie? And it gives you a number of things like form material in the dough excess production speed, uneven mixing, things like that could cause it suggested next steps, you might want to inspect things, you might want to check the temperatures, you might want to look at the controls. Uh you know, you might wanna examine the dough ingredients and mixing process things like this.

So it's got, you know, it's been trained on a large language model that says, hey, what are our most common root cause problems that could result in this case? A cookie deformation in nsgs case, a glass cord break in, you know, brad and braden's case, perhaps it's aaa parts failure or something like that. But we, we, we thought about, you know, how do we train these large language models to make sure that this is really useful locally and we thought, all right. Well, it's not just that it's also, you know, on that next step i need to inspect.

So can gen a i kind of talk me through and walk me through how to quickly inspect this process to see what could be going wrong and to look at the different types of causes that might be uh culprits in this situation. So, you know, it walks through the the data, it looks through the cookie form. Uh obviously, we use our digital twin to look at all the data in concert with each other. And we're using our nice graph view to kind of understand how the process works.

Um and then, you know, we identify an issue and we say, all right, fine, great. I i see that there's an issue but how do i mitigate the issue? How do i fix this issue so possible to fix the temperature fluctuation? Uh what should i do? And very quickly it comes up with a list there of recommendations. But what's really cool about that is we can quickly behind the scenes, inspect the source material, that large language model has been trained on documents after all.

So we can pull up the actual sop for fixing the issue. We can look at that in conjunction with the real time data and we can do that all uh using gen a i um to help us resolve and help us uh think through the issue.

Um and finally, you know, the issue is resolved. So we watch the alarms clear and we watch the uh uh system go back to normal and uh make sure that we're not displaying data that that doesn't need to be displayed.

So, um this will be available to you uh shortly after as a, as an open source project. It's all based on amazon bedrock. Um we have the, the, the source code available in a uh github project. So uh you can build your own shop floor, gen a i assistant. And uh i'm sure what you come up with will be a lot cooler than what we've done with just cookie factory.

So, um i mean, i think the, the key takeaways from today, uh i can't stress it enough with digital transformation. You've got to be working backwards from your business goals, align these business goals, shop floor to top floor. Really think about bridging the gaps between that it and that ot.

You have to build a repeatable data foundation and you have to start with scale in mind. Where are you headed in the future? Um just because we're building something now and we think things can remain on prem doesn't mean that in the future, we won't want to take advantage of things like, you know, generative a i or machine learning or, or things like that. We might also want to do site to site benchmarking.

So uh in the future, we need to think about that future scale. We need to build that into the system. Now, we also have to select the right collaborators. What we mean there is partners. Um you know, we really believe digital transformation in iot is a team sport. We work with great partners like embassy of things. Uh we work with siemens, we work with uh lots of different little isvs.

Um we've got great uh architectures called the industrial data f uh framework with lots of different um isvs and partners uh to bring their special brand of, of uh skills to the table. Um it's so important to pick the right ones. Uh you know, for example, our, our mendex lean daily management solution, a great solution because you know, low code is great for the shop floor. Digital twin helps people visualize. So we love, we love these types of relationships and we love delivering like that.

Um take a human centric approach uh making sure that, you know, we've, we've delivered tools to the right people with the right motivations that are aligned to those business goals. Not just, hey, gen i, i seems like it's cool. Let's try to push that into the shop floor. We have to find the right way to move those technologies and to use those technologies as true enablers of our people and make sure that they can achieve their business goals uh using that technology and finally don't be afraid to embrace the future.

Um building in scale building in, in terms of uh what is the next 30 years look like? What is the next 40 years look like? We should all be super aware of it in this space because we're all dealing with equipment that's 30 years old as it is. Wouldn't it have been nice if somebody back then had said, hey, what about 30 years from now? What should i be thinking about? Right? So you have that ability, you could do that today. Don't forget it. Embrace the future. Think, think about that uh uh the life cycle of that plant and that equipment and uh and embrace that future today.

That's it from us. So to learn more about these new features in sitewise, there's a qr code, uh our new offering from aws and siemens. I encourage you to see their session tomorrow, mandalay bay 8:30 a.m. And uh the final qr code build your own a i assisted digital twin today that will take you to a gi github repository uh where that's all open source. So that's it. Thanks a lot. Thanks for coming.