Using AI to Transform Your Fleet Operations (Cloud Next '19)
Thank, you for coming along to our, talk on using AI to, transform fleet operations, my, name is Mubeen dim and I work in a, Hitachi consulting. Within, the strategy and innovation team, hi. Mandy, Lamont I working over at digital transformation, practice in the UK. So. Today. We're here to talk to you about Hitachi's, predictive maintenance. Platform. For fleet operations, just. To give you a bit of an overview of how this session is going to be structured it's, gonna be similar to everywhere you've been already we're gonna essentially. Tell you why we've. Done what we've done over the past, three or four months so giving you some information around, the industry, issues we're trying to tackle from the fleet open up from, from a fleet management perspective, we're. Gonna spend some time showing you how, we've done what we've done by. Taking you through the architecture and taking you through some of those challenges. And design decisions we've made and then we're gonna show you because I think that's the most important, part right you want to see what, we've done and we'll. Give you a demonstration of how we've applied, this with one of our clients as well. The. Key thing I think all of you are probably wondering is why, do we have Lego and this let's address that elephant in the room you've, probably walked around the conference. Rooms and seen, the Lego at the booth and I think more people have been fascinated by the Lego than anything else essentially. It's a whole lot easier to bring one, of those from the UK than bringing an actual vehicle so, we've. Got this to support one of our use cases around a, are. Guided inspections, and we'll give you a lot more information about that as we go through this session, if. We could hold all questions until the end as well will, allocate some time for that but. If we don't get around to you because we run out of time please, swing by the booth ask for Meredith she'll be able to help you out there as well thank, you it's.
Perfect, I think we've talked about this but essentially, our aims are to show you what we've done explain why we've done it but also show you where we're going next and. Just. To give you all a bit of context, and information about Hitachi, so you've all probably heard of us because we touch pretty much all industries. From, an industrial, perspective to, it nology perspective. We. Are a global technology leader founded. Over 100 years ago and a global industrial, powerhouse with, long-term stability and, deep resources, Hitachi. Has, deep, experience in the industrials. And manufacturing, space and especially. Within the kind of vehicles, and fleet management space, so there's, some interesting information out there I won't talk about all of it but over, 70%. Of vehicles have at least one Hitachi component, and Hitachi. Manufactures. Components. That amount, to up to 45%, of the component, building materials you'll have in a vehicle so we have deep deep. Experience. In the manufacturing, space there but we also as Hitachi, have, a lot of experience, in the, ownership. And operation of, large, fleet, operations, as well so, that's, spanning, from the automotive to. The trucking to the heavy asset, so your your earth moving materials, as well and. Along. Those lines attaches also got deep experience, around IRT, AI and, we're bringing together those with. Our use. Of Google, as our, platform for, bringing our industrial. Experience. And heritage to, the industry, to help our clients within the industry innovate. And move forward. We. Also are a combination. Of 800-plus. Subsidiary. Companies and, that's how you've probably seen a lot of Hitachi. Touch in your world. So. What are we here for today so today, we're. Here to talk about fleet. Management and how. Predictive. Maintenance can, support fleet management activities, fleet. Optimization, activities specifically. Fleet. Management solutions, need. To evolve they need to move from where they are currently. Fleet. Management solutions. Are, often. Reactive. Visualizations. Of the information, that you have available to yourselves. Reactive. In terms of maintenance but, also reactive, in terms of, any. Kind of, optimization. Activities, as well today. Flee, or, flee. Organizations, essentially. Have to compete with a, number, of disrupting. Factors, such, as startups. Other entrants, into the market, rising. Fuel prices and. Lexie mission-based regulations. That's, forcing, them to focus on efficiencies, and within. That. Maintenance. Costs, for fleets account for, 15 to 20 percent of your. Fleets. Total costs so, a prime, area for, optimization. And a prime area for AI and IOT, based solutions, to help reduce that cost. Digital. Technologies. Are. Changing most. Industries, in all areas and this is somewhere where there's, a great opportunity to, provide efficiencies. And cost savings. Just. To give, you all a bit of contextual, background information. There. Are a number of fleet, maintenance optimization. Challenges, that need to be addressed, so, fleet managers often lack the pertinent, asset, condition data they, may have data but, a lot of the data that gets collected across fleets is cluttered, by individuals. As opposed to necessarily. Being streamed back from, the vehicles themselves in all instances. There's. Excessive. Operational, maintenance cost, so a majority. Of organizations from a maintenance perspective do. Periodic. Or cyclical, maintenance, where, essentially you, are maintaining, the asset on a time-based. Process. To, essentially, reduce, the risk associated with that asset failing at any point in time that's. Due to a lack of available, asset condition, data and, a limited understanding of, the relationships. Between the. Assay failure event and the data that you can get from those assets beforehand. Assets. Are becoming a whole lot more complex as well so fleet. Operators, are having to manage. At larger, scales with more complex assets, and that. Brings. With it its own nightmares as well and, there's.
Also. The. Challenge. Around, knowing when to maintain your assets and when, to replace your assets so for example. End-of-life. For assets, is. Calculated. By. Organizations. Usually, on a time basis, we can bring additional, data in to, help those, types of organizations. Retire. Those assets and replace those assets when they need to. Just because a certain time is elapsed. So. This is the maturing, process, the. Move to predictive, maintenance doesn't, happen overnight, we're. On this trend moving from data. Reporting, through, to descriptive. Analytics and, dashboards to help people understand, their assets and through. The use of data and bringing in data you're. Able to move to the predictive. World before, we get to the prescriptive world where you're doing things because you feel they're going to fail or you're, doing things before, you need to do it. Moving. To predictive maintenance is, expected. To save, 10 to 20 percent over preventative. Maintenance practices, and we'll break that down a little bit as we talk through this conversation, and. Why. Now the, big. Question is has, this not been possible previously, why are we able to do this at this point, technology. Convergence is a key reason, we're able to, essentially. Bring, a number of different data sets together from, both. Edge. Devices, within vehicles, within assets, as. Well, as streaming, data from, other data. Sources so we're talking about the Met we're talking about condition, data and all those pieces. Enabled. By cloud, computing, and big data toolkit so they. Really, allow us to stream. And store vast amount of data efficiently. Economically. And securely. Providing. Monitoring. Capabilities. And the ability to make interventions, before issues. Occur they. Are machine, learnings could be trained against these large data sets at. Comparatively. Low. Cost compared, to in. The. Past and advanced. Visualizations, through, the use of augmented. Reality and, also automated. Dashboarding, and those elements can, help surface, the information, to the right individuals, at the right time so, that they can make those interventions, as needed. And. The. Big question we have to ask is well why do you do these things why do you, go. Through the pain of understanding. And working through those data elements and, it's. All about delivering value, for the organization's so fleet operations aren't. Essentially. Anticipated, to, juice costs by moving to predictive, maintenance. Improve. Efficiency, so I'll give you a few examples. If. We look at. Maintenance. As it stands at the moment we talked about it being done on a cyclical, basis if you maintain assets, when they need to be maintained as opposed to for, the sake of it you reduce, the, downtime. For those assets so. For example you don't need to take vehicles, out of service to go and do maintenance against, those assets you. Therefore, make. Those assets available and you can improve your revenues against those at that point you're. Reducing your operational, cost you're not having individuals, having to go and service. Assets that do not need to be done and, you. Also are, able to. Manage. Your capex, and essentially. Focus. And work on those assets that need the work when, they need the work. So. I've given you a bit of context, and bit of information around. Wifely. Organisations, need, to consider predictive. Maintenance need, to understand, their assets better. And. That's understood, and that's a, key. Driver across, the industry however. There, are technological. Challenges associated with this in terms of streaming back fleet, data so. A lot. Of the data that you pull off the vehicles, for example are unstructured. Or, the, data sources that available, come through a time series data, with. Varying. Structures. And schemas. IOT. Data itself, isn't going to help you generate insight, insight. Comes, together from, a combination of, additional. Contextual data sources so, knowing. The actual event a vet events, that were occurring at that point in time, climate. Temperature terrain. Where you were with that vehicle and. Also. Within. The work we've done we've, identified that, you need SME.
Input, You need the, knowledge and information, from those SMEs that are working in that space to, help you focus on the right use cases you. Can bring back data for many sources but unless it's actually serving, a purpose and working, for the use case to help the organization, deliver value, it's not going to be worthwhile. Perfect so, I'll give you a bit of an overview here around what. We have created and why we've created it. So. Essentially, Hitachi. Is used GCP, technologies. To deliver a predictive, maintenance fleet, operational, solution which overcomes, some of the challenges I talked about we. Do two. Things predominantly, and we'll, go into a lot of detail, around this from a. Actual. Feature perspective, and functionality, perspective but just as a high-level. Overview at this point we've, created, data. Dashboards, for maintenance, planners, that look, at the data that we pull out of those vehicles compared. To contextual, data sources and, they. Look. At the full information that we've already created and the AI models, we have created, to, predict, fault. Events and they, surface, the right recommendations. To that maintenance planner at the right time so they know where and when they. Need to maintain assets. Comparing. That against the alerts that are also coming back from the vehicle so it can be dynamically, updated. Throughout. Their kind of planning process but also looking, at the normal, maintenance, schedules, as well for those assets that are not quite. Ready. For this, what. We've also done is working with our client, around, the. Maintenance. And inspection, space as well as we help them identify where, they, need to work on their assets and we're, using an AR, guided, inspection. Application, to help their largely. Mobile, workforce, understand. The, steps. They need to take to complete that inspection against that vehicle so our client has a large, mobile, workforce but, they also have a large. Number of contractors, that they use as well and the. Use of an AR, guarded application, has been super. Useful to, helping those. Contractor. Workforce individuals. Understand, what. They need to do and where they need to do it we'll, give you a demonstration of, that as well so, that is using, the. AR. Components. Of Android, but also using, Google glass as well. So. In, terms, of giving you the overview of what, we've done and how we've done I'll touch on the concept so our, solution, looks, the. End-to-end, using. GCP all the, way from data, ingestion, from, either the. Obd terminal, that we have in the vehicle or additional. Sensors that we instill. In that vehicle and Andy will take us through in a white little while the, Google. DCP, components, we use from an architecture, perspective and why we use those we. Bring through the data and both stream of backs depending, on the use case it's. All around, what we need to show when we need to show it and then. Using Google's. Capabilities, from an AI perspective. Using auto ml, ml. We. Essentially are doing the analysis, against. That data set to. Identify. Those fault events and we're, visualizing all of that information through. Assisted, reality and data, dashboards, using Google Data studio and also. The. Mobile and glass application as I've talked about. So. Just to give you an understanding of, why we're doing what we're doing and how we got to where we are today. Within, Europe a Hitachi. Own and operate a number, of rail. Trains. And we provide those trains to, organisations. But also they, do the maintenance. Services. Against those vehicles as well as. It. Actually were creating those trains they were looking. At what data we would need to pull back from those vehicles and structuring. The development, of those and also. The development of the data capture, elements on those vehicle on those trains to bring back that information so, we could do things like asset. Monitoring, and and, asset. Optimization. That. Has evolved as we've moved on to the GCP, suite that's where we're looking now with, our fleet clients, which. Will show the demo for to. Do predictive maintenance using, GCP so bringing in all of that information is great and visualizing, is great but we are taking that next step forward bringing.
In Contextual, data sources and. Identifying. Those, types of failure events that require intervention. So. Big, question is why ai why, now and the move from kind of descriptive, analytics, to, predictive analytics, is driven, by a maturity, in AI, technology. So this won't be new to a lot of you in the room but. As. Businesses. Are evolving, faster, than ever data has to keep up and deliver the insights for, those individuals, within, the organization's and bring. That data to the frontlines to help them make the decisions as and when they're needed and. Capitalizing. On data is not new how, people have been doing it throughout time the difference, now is that, this evolution, of. Compute. And this evolution, of scale, and speed and, diversity, in the types of models that are available and, the type of Big. Data technologies, that are available through, Google Cloud from, a capture, all the way to a storage. Processing. And insights, perspective, are, pretty. Much ready at this point in time to help deliver those insights, it's. Not about big or small data it doesn't matter how small the dataset is it's, about how you leverage, it with the insight, from the, industry, for, the right use cases to deliver value to those individuals. As well and. This. Is all about the move from understanding, or hindsight. Through, to insight so how do you get. Insight, available, within your organization, to know where you need to work and the, overall overarching, aim is to getting to foresight. Where you're able to go and do things before they occur. We've. All see Minority, Report right. And. Just. To touch on this as we're talking through, this, Hitachi. Has, deep, experience with in the maintenance. World Hitachi, has deep experience with in the IOT, world as well so. We are able to leverage not. Just the AR components, of Google we're, also able to leverage the experience that. Hitachi has across. The, maintenance space from a disparate descriptive. Analytics space, to, predictive, and also, some of the newer things around prescriptive, analytics, as, well so, for example Itachi has a, our, solution, cause that we were able to containerize, and bring, into the Google, environment to. Do certain. Things such as operations. Risk or so driver behavior and we've used some of that insight, not, necessarily, to detect. Those faults but to support that, data set that we using for, the brakes example, which I'll take you through later today and. Also there's, really important. Things there around remaining, useful life, estimation. And. Fall, predictions around. Other use cases that we haven't necessarily touched, on in this space but, the, purpose of me showing this and explaining this is to say we. Have great, experience, in this space and we're able to bring those with, whatever use cases people are looking to work on as well and. All. Of this is in the pursuit, of total. Operation, optimization. And automation, so, it's, about moving from the.
Hindsight. World to insight and also to, the optimization. Space where AI is not only, helping, you understand, what you need to do but orchestrating. And run it running, that for you as an end-to-end system, we're. Not quite there but that's where everybody. Should be fair walking. Towards as well. So. I'll stop there for one second just do a quick time check with Andy how we doing okay fine yeah I'll. Let you know that I'm I, get, told that I like to talk a lot so just, have. To use Andy to keep me honest I'm here to throw him off the stage okay. Perfect. So before, we actually get in and start showing you a bit about the architecture I'll describe some of the features of what we've been doing and when, we do the demo I'll explain some of those pieces as well so, our. Fleet. Operations platform. Does the, normal condition, monitoring. Element so you bring through the data you show some of the basic elements like Mars driven where the assets are and those elements we, also do advanced failure prediction so we have two, use cases we're talking about today one, around failure, for air conditioning, units on fleet, vehicles and one, around, failure. Prediction for the, brake pad, or brake elements, of the vehicles as well we. Also have anomaly, detection models where we look at peak performance versus. Normal performance we. Bring in the intelligent. Driver score elements that we talked about and we, surface, all of this to our end user being, the maintenance planner through, prioritized, maintenance, dashboards, where we offset the. Real-time predictive. Recommendations. Against, some, of the cyclical, work that's been going on as and. We also surfaced that information, through mobile. And glass applications. To the engineers, who, are out there doing the inspections, and completing that work. Perfect. I now. Hand over to Andy yeah. Thank. You very much, he'll give you an overview of some of the architecture, components, about what we've, been working on cool, I so. Yeah we've obviously talked about the fact that the. End-to-end capability, we looking to use GCP for this we've. Built the platform, with. A number of the components so again I'm just maybe just give you a bit of an insight into. The. High level platform. Predominantly. This is best around Google's. Blueprint. For IOT you probably see lots of diagrams, in this area this. Is very summarized, I think, the important aspect in all of this is not you know this. Architecture, doesn't, necessarily for every use case we. Have to bring in and use different components, of the GCP. Family. To, address some of the challenges so again depending, on the speed that you need to consume the data whether, you're looking at these things in real time whether you want to do things in batch. You. May end up using different products this is just a I guess the common ones you would tend to use in such a solution. From. Our point of view I, guess more, recently cloud a cloud IOT. Core is something that we've, introduced. Some. Of our clients historically. Wouldn't have used this they would have used direct. Connections, or. Stored. The information, centrally. But now with IOT core we can connect directly to these devices and again it's part of the platform we've we've proven some of that technology out. So. This this is here - I guess predominantly, we want - this is not just for the sensor data, I think asthma beans mentioned previously, there's. Lots of other information you want to bring into the platform, to. Actually enhance the. The, prediction. Models and. Again. Different. Clients I've got different volumes of information so I think back.
To One of the other slides you, know from, a rails perspective, if they. Have a lot, more days of potential than what you maybe have on the car so, you've obviously the platform you can scale out you. Nabal to process it you're able to. Use the power of the of the GCP, based. On the input. The data that's coming into the solution I. Think. Again, what we've touched on previously there's. Lots of talk about containerization. And again, coming back to the. Algorithms. That are the solution cause it actually likes the colon that we've got we. We, can obviously bring those into the platform so, again we don't always have to use, GCP. Products we can introduce our own technologies, and things, that we've invested, in previously, and developed, over a number of years so we also kill include those and again we'll, see some statistics at the end where we've actually done that and we can solve sure, where the benefit of using the platform. So. You, know we've talked about the back end you know as I say lots of components from a visualizations. Perspective. We've. Utilized, data studio to. Do the the, what we would cluster the standard dashboard capability, so. Again. From a maintenance, planners. Perspective, that, is where we are surfacing, the. Consolidated. Information so. That's where the, person is and making those decisions. To. Support the the future, I guess the maintenance effort is required. And. Again Google. Glass. That's where we're doing the the guided investigation. And again all the, components. Interacting. With the, GCP. Platform. Taking. Data in, giving. Data back after, some of the investigations, have happened I. Guess. These are some of the differentiators, and we can call out from GCP obviously from our perspective, and. To end capability, so. We can obviously collect the data process, it analyze. It and, then obviously do the recommendations, off the back of it that's obviously, very important, I think, as as we mature within, the area of predictive maintenance having. The ability and. I think probably some of the announcements are correct from Google this week you know they're continuously. Investing. In this area of machine learning as well as what, hate actually doing so we can bring our. Enhance. Capabilities, and, use obviously Google's enhanced capabilities to progress that as we go through.
Security. Is a big thing I, think. You know Google is best-in-class from a security perspective a, lot at the time when we you know you could. Figure in the platform, it's just there it just does these things you, don't have to worry about it and there's, lots of you. Know out the box capability, that we get with the platform and, again obviously we've talked about. Different. Clients require different. Needs. Speed. Performance the. Scalability. Is obviously very important. For. When we do a lot of the work around the algorithms. Again. I think we've probably touched on this this previously, but you know my background is bi and it used to be very humid, you would go and gather some requirements. From it from an end user you. Would write you know you'd write write a document, you would then go away and build it and I think in this day and age you've. Got more parties, involved. With the whole process data. Engineering, is very key you. Know death, is no good if it's not meaningful. Clean, so. There's lots of interaction, with data engineers, and the data scientists, but, obviously, the important, people on there is the business operations, people the subject matter experts, who. Understand. How, these lovely. Objects, work and. They need to provide that input and guidance and. Again I think that's we, found that it's very critical. In key to, making sure that anything your developers get gives, you the the, right outcomes. So. Why attaching, Google I think we've touched upon it previously. Attached. Has got heritage within the, industry. Space. And. I. Think probably is a single tagline it would it would probably you, know we were using Google as a platform to, bring our I, guess industrial, expertise, to the market you know we we've. Got some we, can see the power in the benefit of using GCP, when, we develop developing, out of these solutions. So. Yeah, that's a very brief overview of the of the technology I think as I said we're now going to just go through the actual the, demonstration, of the, application. Thanks. Andy, now, I'm assuming all of you have already come past the booth and seen the demo so this be the second time you've seen it yeah I'm, sure I'm sure that's already happened, so. Just. To give you an overview of what we've done with our clients so obviously our client doesn't just have Lego vehicles, Lego. Isn't the enabler for AR to be a success, and driving site it's just an abler for us to demonstrate some, of the pieces we need to, but. Essentially our client. Was. Having is having a lot of work that needs to be done from a maintenance perspective around, to, specific, use cases so eith AC, so air conditioning, HVAC in the states and also their break elements, that's where they spending a lot of their time in terms of maintaining their fleet of vehicles and. These, fiscally uh vehicles is dispersed, across a large. Geographic, region and, within. Their. Organization. They have a number of mobile. Workforce users, but they also leverage, a, contractor. Pool of resources when they need to so, we started, to think end to end about how we can provide a solution to help them both, understand, where to focus their attentions, and when to, create the right interventions, before assets, failed but, also how. They. Could get the contractor, resources, the right information when they needed it visually, as well because the last thing you want as a worker. Is lots of paper. The key thing around all of this as well is. From. A. Data. Kind, of management perspective our. Client, has access to the data that comes off the vehicle so in the Europe, there's an obd to terminal, that sits inside a vehicle that, streams back lots of information about that, vehicle now, a lot of that is just noise right you get lots of information but you need to figure out how, that is useful, for you so, I'll, just quickly touch on one of the use cases we'll talk you through today around, air conditioning, so our. Client was noticing they have a lot of, issues. In, terms of client, feedback for the vehicles that they're borrowing around a/c performance, issues compared.
To A some of the other issues some, of the other maintenance elements, and a lot of their maintenance work was around, repairing. Or servicing, these a/c units, across their vehicles, as and, each, vehicle obviously slightly. Different unit slightly different, maintenance. Shed your pattern, that they have to follow so. We worked with our client to understand. What data points would be useful to pull back from the vehicle around the, AC usage, so looking at the time it took to get from start. Temperature, to, the. Desired temperature is a great starting point in terms of both cooling, and heating but, then we looked at failure. Information, we could gather from that client to their maintenance records also the external. Data. Sources that are required, to support, their AC use case so how do we bring in external, temperature information how do we at, least test, the value of putting. Additional sensors, within their fleet vehicle so that they can have a robust, data set to be able to get. To a fault prediction perspective, what. Our client also noticed, and and has from an SME experience, perspective is an understanding. That if you, get to these eight, air conditioning, issues in. Time, you, can you, can actually remedy, them before they turn into total, failure so for example, if. An asset, has. Degrading. Performance, from an AC perspective, and our model is flagging that and, we know that that asset was serviced let's say within the last six months it really shouldn't be having any performance, issues that. Is probably an asset we want to prioritize getting, to to work on because, those. Issues within an AC unit are usually down to pressure, eroding, out the system coolant, leak those, types of pieces and they're able to identify. Based. On what we've created those. Elements in those areas where you could get to it and repair the hose or and improve. Whatever, issue there is with it so. The whole unit doesn't fail, so. We'll take you through that use case. We. Have, two. Focuses from this perspectives their data dashboards, which Andy will talk you through shortly where, we're surfacing, all that information, and we'll talk you through that and we also have the guarded AR inspection, piece for, that mobile workforce as, well. Andy. Thank. You. So, just. Looking, I guess from a dashboard perspective, this is very we, play class more the bread-and-butter. Bi. Type stuff with. Them means to you. Know surface. The information, and we need to take action upon so. Again. From a. Dashboard. Perspective, what we're doing here is we're effectively. Summarizing. The events. That have happened now those events may be that the system itself is generated, or actually, in this case is where we've, executed the models and. It's come up with these recommendations. Of, the things that need to be actioned, straight. Away so, again this. Is but this is more from I guess work from Earth maintenance. Planning perspective as, opposed to the individual. Again. We can obviously drill into the detail of that we can go and see the, specifics, of where this. Thing is failed so in this case air conditioning, and again, based on so. These conditions cause obviously things that we this is the score that we generate, again. That's based on the the, input factors, for the different, components. And models that we're predicting. Against, so. Again just a visual representation, of the, information, that. We're then ultimately going to go and do some inspection, against, and. That's how we're other beautiful, car.
We. Sort of said that we've, we've talked about HVAC, in and braking. But. As part of the the data, set that we actually had the. Data, scientists. Were actually. Started. To I guess evolved that if you like and started to look at things around drivers. Behavior, because. That's, obviously a key, aspect into, potential. Degradation, especially, things our own brakes where you have a drive. Who's. Continuously. Braking. And driving I guess, not to the optimum again. Something. It's very useful, within the whole the fleet, for. The industry's perspective is, it's also used on the trains it's similar sorts of things where you want to try and optimize, that and again we've built. Out some, capability. And. Again these dashboards here are just showing the, I guess the aggregated, views and some of the the key metrics, that people want, to capture. When, your stockin starting to look at, driving. Behavior and again we can drill, into that detail. Skip, through these but again it again, it's an education, thing so we saw we. Can obviously use that from a where. We have to go in actually inspects and service the vehicle but we can also do use these things from an education. Perspective because. Obviously it's the same driver. It's. Causing the similar sorts of problems and we can obviously guess take corrective action, and which ultimately will, obviously improve that the, vehicles, he's driving or she's driving and just to build on that as well so our clients, now looking at the data we're bringing back from a breaks, perspective, and drivers core perspective, to say well that's good for maintenance but how, can we start to surface some of that information to, that end user that driver to let them know that they're not driving optimally, they're accelerating, harshly or braking harshly not, as a way of telling them off but as a way of helping, them understand, their performance, we carry health. Monitoring, applications, and it's just about seeing what, you're doing because you may be let's. Say naive to the fact that you're not driving efficiently. Cool. Thanks sorry, okay. Okay. Perfect. So this. Is the part where you actually see the. Lego. Porsche is not just here to look nice so, essentially, in this scenario we've, created an application for, our clients mobile workforce. Client. Wants that mobile, workforce made up of contractors, to be able to execute the, right activities, from, a maintenance perspective, there's. Two or three things we're, doing here and this is all about the inspection workflow. Within. This one, I'll talk the, examples. Through as we've got the video playing in the background but, essentially, within, the maintenance world, you. Would be surprised but there are a large, number of times where somebody goes out to, work on a vehicle or an asset regardless. Of if it's a fleet asset or any other asset and they work on the wrong vehicle I.
Come, From a background in, rail slightly, where that, was happening everybody. Feels like they know where something is you get the information from the data base and it tells you it's in that location it's a fixed asset you've been there a million times you get there you see the checklist you, know what you need to do, you do it your market is done you don't realize there was actually an identical, unit right next to any worked on the wrong one so. We use AR for, our visual, identification. Through. Google to help we use or understand, that asset now with vehicles it's easy because, we. Have license, plates we have registration, plates for those vehicles and. With our client they have three different types of vehicles so we've trained it against those three vehicle types but also the license plate as well to remove, that chance, of error for the demo purposes, you'll see today we've retrained the model and we've done, made some of that data for. Unsanitized. It so, you can actually see the AR image. Recognition against. The. The. Lego unit, itself the. Other thing we do using. The application, is we, help the, user understand, where to focus their attention for this inspection, so for example I may, be a add. An AC inspection. Engineer. However, I may have not worked on all, the cars that have ever existed in the world so it's about surfacing, the information, for, that vehicle and it's all about us then training the models to recognize, our clients vehicles. And their assets so scalable. Outside of just vehicles, and and fleet, assets but, immensely. Powerful to helping people understand what they need to do so, in this scenario, we're. Essentially the, p-side Walter out the model was highlighted an AC efficiency, issue this, could be due to gas regulation, within the system or leak. Somewhere the, AC unit was serviced quite recently so it's actually something that we're looking for our client to target and get them out to complete, that inspection and we. Will go ahead and complete, an expression against that unit so. This. Is where we're hoping the magic works and the video will play I, will. Probably, pause, it from time to time just to recap them some of the things that we're doing this. Is the same overview. Of the. Use case and scenario that we're working on but, essentially. User. Is able to see the job queue the work that they need to go and do and. Obviously, they get a lot of information through, from, our, model, so whatever we hold in the model will surface the right information to them about what they need to do where the asset is. And then we'll keep it playing we. Essentially help them navigate to, get to wherever they need to do the work on so for example using Google Maps to get them to their location. For, simple purposes we've simulated that for this part we, get to the asset and they need to identify the, asset so piece. Are talked around earlier around working on the wrong asset is extremely. Powerful we. Need to, make sure that we fix things right first time on the right asset to play. This, is where you'll see the AR element. Come in so the a of recognition, element the. User. Hovers the mobile. Or glass, device, across the asset and data is pulled back from our database. To show the. Pertinent. Bits of information that we want to surface for that use case and for that user for this demo, purposes, we've pulled back the mileage we pull back the air condition score and. The focus area as well now. As we click play the. App. Also provides information to, important, for. That user so that user understands, what it is they need to do so. From an AC, inspection, perspective, there's usually three things you go look at the condenser the. Pressure that's within the system but also check. The fixtures and fittings around the hose is there any leakage happening in that space so. Essentially the users given basic information, and then they're guided through this so if we click play start, inspection, they'll. Be able to that. We've told obviously where they need to focus their attention they'll. Scan, that asset so this is using the camera to pull back visual. Information, check, against our model and then it pulls, back the information, around what it is they need to do so at this point we've called out the three inspection areas now for different vehicles different cars you'll.
Have Different areas to have the hose or have the condensing, unit and, this allows them to click in and see what it is they need to work on and the, steps they need to carry out as well so it's removing. The end of job checklist activities. And it's focusing, that user on what they need to be doing when they need to be doing it this. Can get as complex, as you want it to be dependent, on every. Cause it just they're handy, complex. As you want it to be depending, on what. Asset you're working on so, it's, really powerful. And then obviously the information having. A press PLAY sorry it's, pulled through and that passes, through. To the, database, so that we can have positive feedback around, what was completed, whether, there was actually a fault and it's reinforcement. With reading codes so, our model can learn as we go, so. That's the demo from the AR perspective. And. This, is where we're hoping that it works and we can go next, again. The value of why we're doing this so the value of why we're doing this for our client is to help, them reduce, the. Maintenance burden and improve efficiency from, a maintenance perspective so, go, and maintain those assets before they fail so you don't have to spend more time and result and and, effort, when those assets are out of commission later. On down the line. Improve. The uptime for those assets so if, we can get to a vehicle and do a small repair as opposed to having to a large replacement, it's been available to be used by other parties and. Reducing. That capital, costs so you do not want to be going out and repairing a/c units as frequently, as, you as you, currently do you want to be able to maintain. Those units to a higher level of efficiency, so. That they last a lot longer as well. Perfect. So. Key, takeaways, I know you're, all probably wondering so. What and how much all of this would, factor. In from a data processing, perspective. So. This is just I guess an example, of I guess why the, platform, is, really useful, it's. Part of one. Of the air. Conditioning, investigations, we did some analysis, and we, try to extend. The dataset that we had to simulate. Over. I think 12 months worth of information, so. You can see there were those big numbers you know if we were to try and do this I guess, on premise we probably, struggle, so. This, was a job, that we executed, they, obviously spun up as you can see quite a large. Volume. Of compute, took. About 22, minutes to run and it only cost $12. So. You. Know when, we give this of ability. To our data science team that they're very happy because historically they're they're very I, guess. Mindful, of spending lots of money because. They know how much people were cast but when they were actually amazed when, the. They saw the results of actually how much these things cost. Perfect. And. In terms of the key takeaways from a, let's. Say learning.
Experience, With our client so fleet. Optimization. Is a continual, effort so it's about understanding what. You want to do where you want to do in what you want to try and improve from, a use case perspective, so ACN breaks with the start and our clients now learning all the other things that they might, want to do based on the data we've been able to pull up and surface for them. Data. Isn't enough, so the key, point is as much data as you can bring back isn't enough you need to combine, that data with contextual information but, also with. That subject matter expertise, and the. Ability, to add new sensors, and bring in additional. Data points that you didn't originally, capture, because it, can benefit, what, you're trying to prove so for example with, the use case around brakes we brought in the additional data around, driver. Behavior because it's, very likely someone's, driving badly brakes, are going to be impacted it's, not this sole factor. For why that scores calculated, but it's a combination factor, then we bring in things like terrain, weather do, those have an effect on how people are using the vehicle from a brakes perspective, and. The. Other key takeaway from this is the speed and scale that you can move out from Google's perspective so you have, all of the access, from. Edge. Data, capture, to authentification, of, new devices using IOT, core to, the use of ml and containers, to do the analysis, against that data set but, also robust. And powerful, visualization. Tools to help provide. The information for. That end user to, help, them complete their job so for example, when, we were talking about the guided, a our investigation, application, we. Tried it in two ways we tried it initially, with just. A mobile application for a user and when we tried it with the AR piece and side by side and the, client was noticed. Also that the, people who are using the a our application, will pay more attention to the work they were doing now maybe that's because it's new novel technology, but, maybe, it's that little nudge that you need to in the work that you're doing it's, so easy to say well I've done this form a million times I know what I'm doing if you're having to use the camera and those elements it, focus your mind on the asset that you're working on as well, the. Other part here is all about technology. Is important, but if technology plus the business context that helps you deliver success, and deliver insight, your. People are your strongest asset this. Whole process is, all, dependent, on using, AI advanced, analytics. Through. A change journey with your client understanding. What value you're looking to it unlock. Understanding. The, focus. Areas that you're looking to prove value in and demonstrating. That value to bring other people into the fold and keeping, those individuals. Involve throughout that process either, through visualization tools, so soft, prototypes. That. You can work with them to help the business understand what is coming and. Adjust. What, is going to be delivered in an agile way so, that what they end up receiving, is actually beneficial to what they're doing. Perfect. I think. We've elapsed. Towards, our time and I think we have about five minutes left for questions if. Any of you are interested in understanding a bit more about this, by.
All Means do you take a picture and scan that QR code, or, also. Have, a look on the Hitachi Google website and you'll. Be able to access our. Information. From the applications, that you saw as well and, thank. You very much for coming along today thank you.
2019-04-20 12:53
