The Power of Prediction: How Siemens Helps Automotive Companies Stay Ahead - with Anton Siering
Show notes
Welcome to the Trend Detection podcast, brought to you by Senseye Predictive Maintenance – which gives you visibility and insights into all your assets, from single machines to full plants to help you reduce downtime, increase knowledge sharing and accelerate digital transformation across your organization.
You can find out more about how Senseye Predictive Maintenance can reduce unplanned downtime and contribute towards improved sustainability within your manufacturing plants, by visiting: www.siemens.com/senseye-predictive-maintenance
Transcript
Welcome to the trend detection podcast powered by Siemens. And now your host. I am your host, Sia, and today we are diving into a hot topic, predictive maintenance, and the tech that powers it. From AI analytics to service contracts, how is Siemens helping automotive companies reduce downtime, cut costs, and improve performance? Joining us today is Anton Zirin, lead business development and sales for digital portfolio at Siemens. Welcome, Anton. Hi, thanks for having me there. Great to be there. So before we dive into today's topic, would you like to start by telling us a little bit about yourself and your old Siemens? Yeah, yeah, sure, sure. So I think there are two major aspects to my position customer focus and digitalization. Being a part of a vertical management team, whereas a team have a clear focus on vertical customers, so different industries and different customers within those industries, specifically in the automotive sector, we have been working for the past several years. Um, being responsible for digitalization, which is the second aspect of my position, I'm in charge of tailoring our sales activities for digital services actually towards vertical specific needs, towards customer requirements, so to say. This means I work closely with automotive manufacturers, customers, um, OEMs to understand their unique challenges. Whereby those challenges, they might vary from production line complexity to quality requirements, um, and then, um, by the end of the day, we need to match our digital solutions, um, in order to solve those specific pain points and challenges and requirements. Oh wow. So you are really bridging both the business and digital side then. Well, exactly, uh, it is a fascinating role, um, because I get to see firsthand how digital transformation is reshaping, especially automotive, uh, manufacturing, uh, vertical or industry. Great. So, yeah, thanks for the intro. It sounds like you are in the perfect position to guide us through today's topic. And let's kick things off with a simple question. Can you explain to our new listeners what predictive maintenance actually is and how it's different from the more traditional maintenance approaches? Sure, no problem with that. So, predictive maintenance, what is that? Um, how does it work and, um, how, which data it, um, actually uses. So predictive maintenance uses real-time data coming from sensors, um, and actors, um, and coming from the shop floor itself. Further on, um, um, it has AI analytics to identify potential equipment failures. Um, and those failures, they will be identified before they actually happen, which is really great benefit, I guess. Unlike reactive maintenance, which is another, um, maintenance approach, um, Uh, it's, it's better because, uh, reactive maintenance happens actually after the breakdown. Preventive maintenance, which follows a fixed, um, fixed schedules, um, will perhaps not really see the, the, the failure, so to say, before it happens, but predictive maintenance is very dynamic. It helps automotive manufacturers act at the real, uh, at the right time, not too early, but also not too late. Perhaps you can give to me and to our listeners a simple analogy to understand it better. Uh, sure, let me, um. Well let me try to explain this. Think of it's like this. Reactive maintenance is like waiting for your car to break down on the highway. Preventive maintenance is like changing your oil every 3000 miles, regardless actually of condition of the car or the oil or the motor itself. But predictive maintenance is something completely different, right? Predictive maintenance is like having sensors. That will tell you exactly when that oil needs changing based on actual degradation. In automotive manufacturing, where a single hour of the downturn can cost hundreds of thousands of dollars, this precision and timing is absolutely critical. And as we know, automotive production lines are some of the most complex, right? Uh, where do you typically see the most critical failure risks? Um, that's a good question actually, um, and you're absolutely right. Automotive plants are the perfect storm of fast moving equipment operations which last, um, 24/7 and just in time production is extremely critical and extremely important. Some of the highest risk areas we do see around robotic arms, paint, shops, conveyor systems, but also automated presses. So those are the main focus assets for us. And so when one of those goes down, everything downstream is also impacted. That's correct, absolutely correct. Um, there are tightly synchronized, all of the assets, all of the production lines, so all of the processes, um, very tightly synchronized with, um, with a very small, actually, with a minimal buffer of time. Um, so even a, a, a short, unplanned stop can actually, um, affect downstream operations or entire shifts. So for example, in body shops, we often see critical risks in spot welding robots. A single robot failure can halt an entire production line actually, right? So paint shops, another example, are particularly sensitive because they involve complex chemical processes and given those. Complex chemical processes, also a lot of environmental controls around those processes. So even something, um, as simple as a F, um, motor starting to fail can lead to quality defects that might not be discovered until the final injection. Um, and so this failure. Um, can result, um, finally in, in a very costly rework and, and, and quality, um, effects. So predictive maintenance here helps manufacturers to detect very early, um, the first signs of failure, unusual vibration, temperature, spikes, pressure, perhaps, um, and so it gives an opportunity for the operations. Um, uh, for the operator itself to intervene before things escalate into major breakdown. Yeah, that makes total sense, especially knowing how tightly integrated everything is. But now perhaps just let's get more specific and let's talk about the assets, yeah. So how does predictive maintenance apply to assets like robotic arms, for example? Hm. All right, so let's get a little bit technical, right? So each of those assets has a unique behaviors and, um, also unique failure modes, of course. Uh, with robotic arms, as you ask it, uh, we often track things like motor vibration, joint wear, um, overheating, um, and things which are similar to, to the very, um, to the hardware itself. Um. When, let me put it like this, we also monitor more subtle indicators like for example, Changes in acceleration patterns or energy consumption that might indicate some very lean and small bearing where, um, months actually before actual failure. So there are, there is, there is a company, so to say, of, um, of assets which we are monitoring within robotic arms. Understand. And what about conveyors or paint systems? Yeah, this is another application, so to say, or in other assets where we can apply predictive maintenance. So conveyor belts might show anomalies in torque or alignment, right? But we'll also track belts, tension, vibrations, and motor current signatures as well. Um. And paint shops, for example, even minor disturbances in airflow or pump pressure can affect quality. We can monitor everything from spray gun patterns to booth temperature uniformity, sensor predictive maintenance, which is the solution. Um, uh, we can apply sensorpredictive maintenance for sensor data, as I already said, from those systems and use AI to identify those patterns, um, those behaviors. Um, and then applying, uh, um, um, um, artificial intelligence, we, we actually can indicate the early signs of, of the failure as I already said. So sometimes, Um, actually weeks in advance. Um, what's really powerful is that the system learns, uh, the unique behavior of each and every asset in a normal operation, so to say. Um, And can basically also consume its normal behavior in a specific environment. A robot in a hot stamping area, for example, behaves completely differently than the one in the final assembly, and the AI, which is applied and deployed within Sensor predictive maintenance, accounts for those contextual differences too. So in the end it's not just about avoiding breakdowns, it's about protecting quality too, right? But many plants already do schedule preventive maintenance. Why is that not enough anymore? Why should they switch to predictive? Um, that's a very good question. Um, and that is actually true that preventive maintenance is widely used across many manufacturers, automotive manufacturers, um, but it's based on the fixed intervals, as already said, not actual asset condition. That means you're either maintaining too early and wasting resources, of course, or too late and you may risk some failures. I have seen cases where manufacturers, uh, were replacing perfectly good components and just, just because the calendar said so, that's very expensive and wasteful, of course. But sometimes the opposite happens too, right? Yeah, sure. Sometimes we have some opposite cases too, exactly. Um, I have also seen equipment fail two weeks before scheduled maintenance, um, because, um, because the usage patterns, so the behavior within the plant, within the line had changed. Predictive maintenance is data-driven and very dynamic. It adjusts in, in, in real time, so that the asset is. Um, is actually, so, um, we're able to say, um, consume the, the real environment, the actual, um, uh, conditions, so to say, of the asset. Um, the weir, um, the usage, the load condition, and so forth and so forth. So, for instance, if a production line runs over time to meet demand, um, the maintenance needs, uh, change accordingly. You know, um, and the system recognized this, um, the system, by the system, I mean, um, sense of predictive maintenance, um, is able to recognize this and adjust its prediction, basically, right? So it's very dynamic. And that's the key, um, I guess, in automotive, um. An industry where each and every downtime is very, very costly. It disrupts everything from production to delivery schedules and can even impact customer vehicle deliveries itself. Right, and I think those disruptions ripple through entire plants as we know, right? So, maybe let's talk about real world impact now. Can you share with us a real-world example of some automotive company that already uses Siemens predictive maintenance here sinceI in this case? Sure, that's, that's a very good question. Um, so just from the top of my head, one example that stands out is a customer using Sensai who was able to save $45 million in labor in unplanned downtime, um. Uh, and this, uh, on at a very, very single site. So you have to imagine that that's really a number, right? So only one site since 2019 when the sensori was first deployed there, um, they already been able to save so much money in labor, but also in, in, in unplanned downtime. Wow, yes, $45 million sounds, is a massive number for sure. That's true. That's true. That's, that's really is, um, but what's even more exciting is how quickly they were able to, to make those returns. Um, uh, so the return on investment was reached in less than 3 months after the scaling phase, of course. So initially they started with just the most critical assets as the most customers should do. Um, basically the bottleneck equipment that would cause, um, the most destruction, um, in case this will fail, of course. Within weeks, the system identified, um, developing faults in a key press, um, in the press line, uh, that would have resulted basically if it would happen, uh, there was a result in a catastrophic failure during the peak production time, right? So it's really a disaster, um. And by addressing it during a planned maintenance window, they avoided what would have been a multi-day shutdown. So that's, that's great, right? So the confidence. Basically, this gave them to lead rapid expansion, uh, just a simple example, right, uh, across basically other assets and other lines. And now they're using predictive maintenance insights, not just for the, for the maintenance, but also for the production planning and spare parts, um, inventory optimization. So there are also some other use cases possible using. Yeah, yeah, for sure, that's a powerful result. But when we talk about automotive industry, um, we know that it does not work in isolation. It also relies heavily on suppliers and OEMs often work with tier one and tier 2 suppliers, yeah. So, Can predictive maintenance be extended to supply networks as well? Yeah, well, actually it's not the use case, so absolutely, yes, we're seeing more manufacturers, more automotive manufacturers encouraging or even requiring their suppliers to adopt predictive capabilities, predictive maintenance, in general, predictive capabilities. Um, with sensor predictive maintenance, the solution can be applied across plants, so it's horizontally scalable, right? Where it is the OAM's facility or tier, uh, tier one supplier. Yeah, so in the end we can say that predictive maintenance isn't just an OEM advantage, but it can actually strengthen the whole supply chain, yeah, for sure, for sure, totally. This kind of shared visibility helps basically to make supply chain more stable and also reduce last minute quality issues or delays. For example, we have OMs who are sharing predictive maintenance best practices with their key suppliers and um Even providing access to, uh, to the same platform. Um, and this, of course, creates, um, the common language around assets, around, um, assets, health and reliability. When a supplier can predict and prevent failures in their stamping press or injection molding machines, um, it directly benefits the OAM's production stability, right? So everything is basically connected. Um, we're also seeing innovative models where OEMs and suppliers share their cost savings from reduced downtime, creating. That's becoming critical, especially with today's global supply chain challenges. But as we also know, automotive companies are often short of skilled labor. Yeah, what happens if a manufacturer, let's say, doesn't have in-house data experts to manage all this? Yeah, and that's exactly the right spot for Siemens to come in, um, uh, with another, say, solution. It's a Siemens data enhanced service contracts, and those contracts can make a real difference, right? So we go beyond delivering technology and we're combining ongoing monitoring um with expert-driven data interpretation, so to say. Um, and those, this combination technology and expertise onsite, basically, um, is there to ensure that our customers can act based on, uh, on, on data without needing an in-house analytics, um, teams or experts. So basically you're saying that Plant doesn't need to hire any of its own data science team. Siemens can essentially provide the support. Yes, that's exactly correct. That's, that's absolutely correct. Take the case of one of multiple CNC machines in a manufacturing plant, right? So through the service, through those contracts, and by leverage, leveraging the data which might be extracted from the machines, Siemens experts remotely, they don't need to be on site remotely monitor the various machines' production data and health statuses. And if the efficiency, for example, of a CNC machine drops, just for example, say due to local operator faults or gradual performance decline, the system detects this immediately, and it automatically raises the service ticket by a shuffle monitoring services. It's another solution. And Siemens expert steps in to analyze the root case and to help the customer to solve this case. OK, so they basically guiding this fix, right? Yeah, yeah, exactly. So, so instead of requiring, uh, the plant, uh, maintenance managers to, um, interpret the complex data, you know, our experts identify the, the, uh, identifies the cycle time issue and recommend the fix. So very easy. Um, and this can be done remotely or via on-site support. It's up to the customer to decide which, which type of that, uh, is preferable. Uh, we have had cases where, um, we helped customers solve problems they didn't even know they had, right? Because we are deeply into the data and trying to analyze the data each and every time, um, the data is coming. Um, so I said, we're identifying the certain shift patterns, uh, which, uh, um, causing Uh, for example, accelerated wear, um, on specific, um, um, uh, on specific equipment. Um, and, uh, this real-time support is, um, included, of course, in the service contract, right? Um, uh, making it basically even, even easier, right? Making it, um, plug and play model for the customers with limited internal resources. It's particularly valuable actually for mid-sized suppliers who want enterprise-level capabilities without building large data science teams on their site. Yeah, that's for sure a huge value add. And yeah, as I can see, AI is clearly at the heart of it. So speaking about AI, what role does it actually play here? Can the system, let's say, learn across multiple production lines, lines or factories? Yes, absolutely, absolutely. Um, that's one of the biggest benefits actually. Sensor AI learns the normal behavior of the assets and then detects deviations. Yes, OK, so then it's just, it's not just a monitoring solution, it actually learns the context. Absolutely. It goes actually even deeper than that. The AI is able to understand the context. It knows that a robot working 3 shifts behaves differently than the one working on the 2, so the contextualization of the environment is already inbuilt in the solution. Another example is basically summer heat affects equipment differently than the winter cold. You know, yeah, and can it apply then this learning beyond a single machine? Yes, absolutely. Um, it also supports fleet learning. So if one robot welder in factory A fails, the model helps identify risks basically across similar assets in factory B and also C. What does it mean? That means, um, that the operator itself is able to make smarter decisions at scale, not just on asset by asset, so to say, not only on one asset, but really across all the fleet. We have seen this prevent cascading failure across multiple plants, for instance, um, When bearing failure, for example, was detected in one of the plant's conveyor system, the AI flagged similar risks signatures, so to say, patterns in two other facilities, and this does allow. Basically, preemptively replaced during scheduled downtime. And then the system also keeps improving over time. That's also correct. Each and every failure, um, every successful prediction, every false alarm teaches the eye to be more um. Yeah, how should I say, to be more accurate and more nuanced in its, um, in its prediction. Yes. Wow, yeah, that, that sounds really powerful. And now, this discussion actually leads into something we are hearing more about with every day, uh, which is sustainability. Automotive companies are under pressure to hit sustainability goals, as we know. So can predictive maintenance help them reduce waste or extend equipment lifetime, for example? Um, so, absolutely. Um, yes, um, definitely. So by replacing parts only when needed, uh, you reduce waste and also extend the usable life of the assets, um, uh, of the assets too, yes. Very cool. So basically predictive maintenance directly cuts down on any unnecessary replacements, right? Yes, exactly, exactly. It also lowers the energy use, the quality scrap, and also the consumption of the spare parts. We have seen customers, to be honest, which were able to reduce their spare parts inventory by 30%, which is quite a number, right? while actually improving their availability because they know exactly what they need and when they do need that. Mhm. And what about energy efficiency? Does it play a role too? Yes, absolutely. Energy efficiency is another big win of predictive maintenance and sani, especially. …