Duration 30:18
16+
Play
Video

Hitachi applying AI in Industry 4.0 digital transformation projects By Kim Naess, Hitachi Vantara

Kim Naess
Chief Architect at Hitachi Vantara
  • Video
  • Table of contents
  • Video
Video
Hitachi applying AI in Industry 4.0 digital transformation projects By Kim Naess, Hitachi Vantara
Available
In cart
Free
Free
Free
Free
Free
Free
Add to favorites
34
I like 0
I dislike 0
Available
In cart
Free
Free
Free
Free
Free
Free
  • Description
  • Transcript
  • Discussion

About speaker

Kim Naess
Chief Architect at Hitachi Vantara

I am a Chief Analytics Architect for Hitachi Vantara working with customers in Europe across various industries. I am also a subject matter expert for Norways Research Council on AI and IoT, where we review public funding to research and innovation projects. A final engagement is with Norways Computer Association where I am a on the board for BI and Analytics arranging Norways largest Analytics and Data management conference

View the profile

About the talk

Hitachi is one of the worlds largest manufacturers, we will share how we are applying machine learning to gain advantages in our own Digital Transformation journey, and our customers.

Share

So hi. My name is Kim Nason. I am a cheap solutions architect for it's not even tore up. I am here to talk to you a little bit little bit about what have we are doing with regards to our own digital transformation and what we are doing in the in the space and what we want to what he wants to achieve I going forward and also how we can apply this to help our customers. So with that, let me take you on a little journey because when you don't you decided to contact me and they wanted to hire me I was like, well, it isn't that like HiFi systems and stuff

and I didn't really know the breadth of what he taught you can bring and what the breast or what he thought you were doing so that he thought you were a hundred companies. That wasn't something I knew until I actually started working here today. I see the Enormous advantages we have a ring operational technology and information technology together for bringing friends in this case a train my field with thousands of sensors me to explain to you a little bit closer afterwards and how we build it and how we fix the maintenance of it

and how we can build a more reliable train service in the UK. I'm on Bristol sign on Brain Trust in robotics and I'm a huge fan of Tesla and SpaceX and you don't musk and very happy to say that we talk to you also investing and how was it developed robotics everyone develops and our Delivery Systems to autonomous driving autonomous cars. so Back in the days when we used to work with business intelligence. We used to work with Erp systems. We used to work with CRM systems be used

to create a data warehouse and I'm used to create it ourselves schemas be used to waiting our beautiful basura stations and what not and that was all well and good with the transactional data and financial data to a certain extent. That is fun. I have to say it's a lot more interesting when you actually start working out feed system with data from sensors from the machines from humans. You name it. But with the explosion of data coming from sensors machines humans a Tetra we need to start thinking about a transition to be I to AI because a

lot of that can happen unless you have some kind of mathematical algorithm trying to Crunch the numbers for you. If you can get answers a lot faster than you cared. If you are applying more of a manual DVI operation without said, I think it's still a lot of the work that's being done in this space. It comes down to data and how you are being presented without and this case a lot of data that looks like a mess a big mess. I other if you're trying to create a data Lake and you're trying to create it on your you're feeling the Dale a lot of different data from a

lot of different sources in a lot of different formats. It doesn't look good. So Ashley for me I'm being a person has a slight OCD one of the first Start doing is just want to start sorting that data to try to make a sense of it. We still do this. We still need to do this in the next step. You want to arrange that day. I'm finally you want to talk with sending them. Visually I used to do this as a kid. That's today is is a lot bring me a lot of fun memories, but maybe the most enjoyable is actually just taking all of that Legos all of that data and

throw it around play it play with it with your data scientist. Try to make a meaning of the data and try to see if your machine learning algorithms deep learning whatever you utilizing to starts a crunching that data started giving you some answers or you can just train your models into Direction you want to go in the essence of what you're trying to do? If you want to you want to start creating some kind of system and the chaos that you're having. So with that said I

just want to reiterate I think I think I mentioned that detox tea already have about 800 companies in total. I believe that he talked to you as a hundred and nineteen manufacturing sites all over the globe. And every and each and every one of these examples as a see here and in the slide, all of them, we have implemented AI or machine learning over a thousand times in a lot of different use cases. I did industrial scale already. So most likely we are doing it and we're doing it on your windows for your birthday. So we will have

some kind of inside into some kind of experience is how it's being done. I want to talk to you a little bit about the the key use case perhaps this is about 4 years ago for a 5 years ago and Itachi rail felt that they wanted to move into the oldest train Market in the world United Kingdom. Anyone who's seen the the layouts on the on the operations of the trains in the UK knows that is very very complex. The goals that business challenges was they wanted to modernize and improved rail Transportation reliability. They wanted to reduce

maintenance costs and they wanted to utilize technology to find these answers one of the goals that they said about what's to be able to deliver trains as a service. Many of you are utilizing a service models already and I'm pretty sure that whatever Isis Torres mobile utilizing it's not as expensive to set up a this is poking millions and millions of pounds per train for train set and we are scaling up two to three hundred train sets during the during this contract for Great Western Railway on for Virgin, Atlantic. Average ER wait wait,

no Christopher things each and every train are fitted with three to 3500 different sensors that measure some kind of device on the train and feedback data back into a system. What we are very early started to see you was that build a solution. We really need it out there scientists to work closely with the two main exports so we could bring a data scientist from the 6 neurology. I Who would know how to build up the mathematical the machine learning models the main experts to identify

data train those models needed the domain experts to tell us what are the are the mean value is here. What is bad? And what is good? Where do we want to be? If you want to be in the middle do want to be high. Do you want to be low getting all that inside it takes time. So I needed to do this for a lot of different incidences from a lot of different sensors. So Getting up to a certain level of predictability into your model on the ability to predict failure Swensons is

Amir equipment is going to be very critical. I can be very time-consuming. So we were applying them all on U sensor data to estimate the probability of all kinds of different various types. And then we have the ability to control trade off between Rico and a false alarm minimize overall maintenance cost. The architecture that we used to build this was that whenever the trains were operating as you can see in the middle of the picture we had the trains fitted with 3G and 4G transmitters that

they could send data us. They were traveling. We also had happened at certain Junctions or certain stops and a service intervals, of course where they could offload more data on it. This data was usually picks up by Natasha software update integration. That's how we will utilize seeing a difference iot streams. In order for us to do this and utilizing some data ingestion ATL procedures, so we could in put some raw data directly into into I do and we could also utilize how to build it and then feed it back into time series event database

for Apache hbase and analytical database. we were also with a pencil able to orchestrate data streams of Kafka and Now all of this was then used to visualize in the analytical layer visualize the data so that we could build reports and dashboards for the difference rules within he talked to you well, and so I sing a lot of the data being fed from the sensors to build a rule engine to do. Gnostics an event for supposed to do to those with epistick sound machine learning. So all this time letting sue us being able to do

what you do in Seattle at that the last section of the presentation layer. So we were able to use this data visualization to do maintenance planning events definition processors having dashboard and reports rating for management and train operators having a system admin up and ready and a definition editor for the events for the rule engines. We were able to build real-time alerting diagnosis and a trend analysis in addition to real-time monitoring of the day.

So the key outcomes of this was that of course, he thought she was now able to avoid unexpected failures. We are paid for having the trains operating at the soo line for a month per month. So if the if it rains been operating in a since we were not get we would not get paid. So we were avoiding unexpected savior speed increase the availability of the equipment's avoided service Interruption on accumulated over 20 million pounds of savings end means him a call when you lie by adopting

a digital service model and a prediction model strengths to mention of performance degradation using a memo to reduce the sound. Might not be too surprising but I can't disclose that the the two most common failures or Grace that cost to get directions on the trains where the train doors and the brakes. So we was trying to help me to get a reason for that but still those are two of the main ones not a big surprise, but it's good to get that and all the shops out there. This is a this is a story from a Scandinavian use cases Stunna line operating fairies

in the in the between Sweden and Germany Finland and Denmark and in the beginning they were discussing planning optimization and that can be boring on a lot of difference diable. So right with you a lot of the different possible data sources to be with cheating with an AI capsim if you want since that's what they referred to it, so they were not doing this in order to replace the captain or do you have a Thomas ships? But they really wanted to have an AI engine being trained by the most experienced captains of thin line

on how they were operating a vessel a difference at winds current steps. I try to be more fuel-efficient on the and of course have a safe navigation screen. I am at a aye captain. They're newer captains who are just being promoted to Captain Ohio. The captain they can have that experience from all of their most experienced Captain right next to them telling them and I'll be getting both the waters that the winds and four. And I said I said I said align themselves dead planning a trip and having a vessel in the safe and at the same time truly appreciate its craftsmanship

practice makes perfect caption officer could learn how to shoot up some ice quicker. I also wanted to show you and talk to you a little bit about how data is agnostic and how we can utilize data across the different Industries. So we are taking our practice of how we have built a digital transformation Journey for preventing our own manufacturing plants for Apache rail for a lot of difference in travel companies who is building it for clients as well. I love this use case and this is a story about sending Max a gaming studio in the US

and the game is to do that owns. A lot of different game tigers are one of them being the one you see here. I do one of my old favorites for mine. I was in in the in the in Middle School Leyden High School And they had search challenges Cinemax were having five different game studios who all had five different game titles. And that meant that they had data silos. That means that they had issues with their God's the format that they are the data that was being outfitted from the from different game developers on the game on the gaming

behaving themselves different formats, everything from databases and data warehouses to Json files and another They're also having issues with the gods of data integration. They were doing a lot of manual scripting on it wasn't being optimal and the way that they were being unable to unborn you sources on your data Austin architecture. Definitely were not helping them at all and the date of volume because it was exploding especially as to start a new soul. I sing more and more data services starting to creating

massive multiplayer online role-playing games. So what they started to do, what's that? They needed to start analyzing data and I needed to start analyzing data Foster. They needed answers to questions like are certain bath is too difficult to AC when they were analyzing gameplay in a massive multiplayer online role-playing game. I needed to understand this. I need to understand that maybe not letting real-time. But at least when the player or logging off and sending data back to the servers, they need to

understand how they play at reforming so how quickly do the players are turn off the bay log off what Bibles for Missions game is prefer or seek out. Do they have any favorites? So can we pretend to be something can we create a Netflix you a recommendation looking field of when the when the gamers get back into the game. So will it be will it be set on that Quest or another question be tossing? In fact monster? Another one. It's going to be based on your preferences and I am perhaps the demographic of other people are enjoying the same ones that you are so

that we can keep players happy and keep players engaged but we also need to stop thinking about so if the if they are, is there a particular points in the game when the players tend to the battleship. When do they start to really quit looking back in the system so that we can keep them on a recurring subscription model not paying for a month. Multiple Cinemax. I can get them. Can we then recommend them you title some things to be we would expect them to like a perhaps give them sneak previews.

So the solution so you see hear the game studios. They had Tango software arcade bethesta, not ID software and machine games a lot of difference and I'm pretty famous Titus are Wolfenstein you had Doom, of course you have Elder Scrolls and more. Are there mission was to improve customer loyalty and analyze our customers play the game now first off the started to you see that the amount of data that they were generating. I was huge eyes of 250 terabytes per day on

an old numbers that is grown. I need license. That's how it is equation to actually Feed or where I am or crunch through all that data connect to it and then send it off into the data Lake which today it's on Amazon and they utilized start on the street story on the red shift ETL capabilities to build the day the lake to the to the data scientist 2122 to understand the the data in the questions being asked in the previous the previous slide. I'm done at the front and

visualization never utilizes the Taos visualization capabilities, and we're supposed to feeding a blow for a self-service before you If you look more Deep dive into the into the architecture now, you can see that the the h a w s components are there technical strategy. What's the build this with redshift data Brooks Caldera Oracle? And Prince how how it gets the daytime from the Oracle databases or the from the Fastpass TDI than Aggregates the data and create CSV files put the been into redshift. We are copy command. I

don't know. We're also utilizing some of our adopted execution engine who enables picked out actually orchestrates a spark jobs. Which was in a pretty useful for that. So this enabled them to say that that's one access throne and spell less cost. I'm dragging that slain is recorded and analyzed it develop receipt the perfect balance of Challenge and Triumph to key players engaged for years. And in the end the outcomes were that they were able to say that we are we are able to have a better analysis of the data. We are able to analyze up to ten

times more data from give me Behavior then go to be able to do previously allowing their analysts spend more time studying studying gaming Behavior instead of loading data and also reduces time needed to launch new products from months to weeks. They were also able to do more automation golden knights work that had previously been done manually or with it also enables greater efficiency with regards to having better quality of data on one Consolidated view across the different gaming titles Studios. Swing through procedures is

also reading on an increase productivity for inside Center billing and accounting. So I said that also that I wanted to to discuss a little bit about how we are actually. Sorry about that. So how we are approaching some of the projects that we are tossed with either return label and Itachi or mixed only when working with clients. So you like to take our customers on a data Joni obvious me to see that on that data Journey. That's a lot of different obstacles.

Is there a huge growth of data complexity both with regards to the state of being on the edge on premises and different clouds in different formats in different databases? You name it? Will this create major challenges especially if you wanted to to bring a data UPS practice where you want to order mates be the day the pipeline for what's possible sore breasts and foremost. We see that there's a big skills Gap with big data tools Big Data project where you were building Cloudera hortonworks environments

on-premise. I'm a lot of the times I T gave that job. So I the database administrators of storage administrators on the big fan of there was that neither storage ottoman or the database admin, where are usually very proficient with Java, which is a requirement when working with potato. Also, it's still a heavy it moments. There are a lot of different things that needs to happen and it being involved in an it. Can then sometimes it's because he wants to have their processes done and ready

and coming back by way of working isn't always too easy. The date explosion. I think I mentioned that I was still a lot of unstructured data and the ability to actually get all that data and start I would like nothing more. Let's just arranging them and sorting them before you can unlock them as hard and you need to start stinking bacon order to receive this data Practice still 80% from this project. That's the case that should be sometimes created data silos as well as we've seen statistics saying that you

have a time clouds / organization. An Indian deploy machine learning models can be hard and you don't always have an automated process for updating Tomatoes when you put them in the auction so you can train them and retrain them all new data. So what we are trying to do is that we are trying to take I will come over at customers on a journey on a digital transformation Journey. We are helping them and nibbling digital experiences. And in this time of year that we are in now and the in the year of endemic we see that a lot of different companies are scrambling

to give digital experiences on a lot of different levels to their to their to their clients. We are helping them Drive operations Excellence, you're helping them modernize the digital court and all of this is needed to become a data-driven organization. Because it can't really be done unless you start counting date on Hunting data that can answer the questions in your organization. Now this enables duration Innovations digital animation. It can then help them to find their strategy

it can Link Technology to outcomes. You can augment their existing Resources with business architecture planning digital centers of excellence data science Engineering Services. Sorry about that. So both internally. Actual clients. We are delivering the best to attack you supermodel software. It's all data is any Industries any Cloud hands every innovation? we are here to help both on our own companies and other clients succeed in any way we can I think you can have some. Okay, I don't have covid. It's just my asthma that sometimes

picked up. So my find my final two slides are are are here. So first off stop thinking about leveraging data start thinking about data monetization have a plan for data governance to both find a leveraged all of the data that you have in your organization. Then try to understand what kind of car you have to ascend to the station of attention Administration Montessori application of knowledge this increases the information in confidence confidence and reliability while enabling it coming

Not any sakshi. We are doing this. I'm doing this since only Onyx when we first started creating a world map with our customers. We build scrum teams that create a PSN user stories to fill the backlog. Let me put this into a development practice, but we plan design develop test and put it in a demo and then be released to production during the production. We have a testing face as well. And then it sends to final production. I'm hopefully at the minimum viable product we can reach right? That's why we can create that service better

and better all the time. So with that said, I think I'm running out of time. This is our full amount of data management portfolio. And my final say is that we are powering good. One of the missions are Petoskey is to drive social economic and environmental value. So what can we do? What can we do together? So with that I'm going to say, thank you and handed over to Lena.

Cackle comments for the website

Buy this talk

Access to the talk “Hitachi applying AI in Industry 4.0 digital transformation projects By Kim Naess, Hitachi Vantara”
Available
In cart
Free
Free
Free
Free
Free
Free

Ticket

Get access to all videos “Global Artificial Intelligence Virtual Conference”
Available
In cart
Free
Free
Free
Free
Free
Free
Ticket

Similar talks

Sudip Chakrabarti
Venture Investor at Madrona
+ 4 speakers
Ian Foley
Venture Partner at Level Ventures
+ 4 speakers
Anupam Rastogi
Partner at Emergent Ventures
+ 4 speakers
Vik Sasi
Partner at Dreamers VC
+ 4 speakers
Dennis Liu
Investor at Ford
+ 4 speakers
Available
In cart
Free
Free
Free
Free
Free
Free

Buy this video

Video

Access to the talk “Hitachi applying AI in Industry 4.0 digital transformation projects By Kim Naess, Hitachi Vantara”
Available
In cart
Free
Free
Free
Free
Free
Free

Conference Cast

With ConferenceCast.tv, you get access to our library of the world's best conference talks.

Conference Cast
561 conferences
22100 speakers
8257 hours of content