A giant company built on tiny components: Interview with Bosch industrial internet boss
In this interview, Stefan Hartung, a senior member of the board at Bosch, talks extensively about the industrial internet, detailing some of the components and devices the company uses to give old machines a new lease of life, and provides some insight into the company’s plans going forward
Bosch is as relevant in today’s computerised world as it was after the end of the first industrial age, and the company’s main concern now is keeping it that way.
Its relevance comes from making the power tools and household appliances most readers will be familiar with, and also from its development of ideas and technologies which are likely to shape a future which many of us haven’t even thought about yet.
Nowadays, all the talk is of Industry 4.0, an umbrella term to describe a range of technologies which have at their centre two tiny components: sensors and chips – both of which are Bosch’s essential stock in trade.
And if you want someone to blame for the Industry 4.0 phrase, look no further than Bosch, because it was part of the working group of German industrial giants which coined the term in 2011.
In the intervening years, the description has started to make sense, as it refers to the simple idea that the world is going through a fourth industrial revolution.
Just to provide context to the idea:
- the first industrial revolution was brought about by mechanisation and steam power;
- the second industrial revolution was essentially about electrification and mass manufacturing;
- the third industrial revolution is said to have been led by computers and automation systems; and
- the fourth industrial revolution is where everything is connected together into what are called “cyber-physical systems” – meaning the integration of software and hardware and even humans.
Such things are probably more important to a company which has been around as long as Bosch, but maybe not so much to companies which emerged in the last decade or two, for whom “100 BC” might mean “100 years before computers”.
Which is about the time when Bosch was established – 130 years ago to be exact, and computers of course only started becoming widespread about 30 years ago.
But in an interview with Robotics and Automation News, Stefan Hartung, a senior member of the board of management at Bosch, says this is just the “first phase” of a new learning curve for the company.
Hartung’s responsibilities include industrial technology, energy and building technology, manufacturing co-ordination, and general strategy.
The company is building a new, billion-dollar semiconductor fabrication plant in Dresden, Germany, and is preparing to launch an artificially intelligent computer for vehicles to support its advanced driver assistance systems, or ADAS, which is, essentially, autonomous driving technology.
ADAS is a big money-earner for Bosch, which says it earned more than a billion dollars in the sector. Industrial internet, too, is big business for Bosch.
The Bosch IoT business also brought in a billion dollars for the company. And, according to Hartung, it remains a “giant opportunity”.
Stone Age nostalgia
Research by the leading management consultants all show that the internet of things is potentially a trillion-dollar market in the next decade or so, and the industrial IoT will be one of the main drivers of that growth.
For now, however, many in the industry are still on a journey of discovery, learning what IoT can do for them, in their factories, especially as they often have legacy machines that date back decades.
But while they may consider old machines a problem, Hartung says they are no barrier to connectivity – it can still be done.
Hartung says: “A large part of industry is in what you might say is a brownfield state, which means it needs a lot of development, in terms of connectivity and in terms of Industry 4.0 standards and applications.
“But we showed at Hannover Messe that it’s possible to connect machines and equipment and what you might say are ‘Stone Age’ equipment to the industrial internet of things.
“We put this old Bosch lathe on our stand at Hannover Messe, and you don’t find many of those in industry right now, but you find similar kinds of situations where you have machines that have old controllers which are not internet-capable. So what do you now?
“The good thing is that you can put sensors on these machines very easily and, in the future, even more easily, especially with low-cost MEMS [microelectromechanical systems] sensors and other sensors which will give you data about the machine operation and the manual operation, which is linked to the machine, unless it is unlinked to all the actual control processing which is in the machine equipment.
“So it’s not necessarily the case that you need to exchange the entire SPC [statistical process control] controller and PLC [programmable logic controller] and so on. You don’t necessarily need to do that.
“You can just put a wired or non-wired sensor on the machine, measure acceleration of some axis and the vibrations and sound and other things which then tell you if the machine is operating well and what it’s doing correctly. You can analyse what is actually happening.
“With sensors, you can very easily sense if drives are running, if the machine is operating, or just idle. And that gives you exactly the kind of realtime data which you need for Industry 4.0 standards.
“What really differentiates Industry 4.0 is realtime data from the location of the manufacturing up to the business information data suite which you have running the company.
“This is a complete contrast to what is the classic manufacturing process, and also to some extent the lean manufacturing methodology, where you always planned what you want to do, you had these production lines, people were working on those lines, and you only analysed afterwards the data relating to the production output.”
Second life
Like many industrial companies which used to concern themselves almost exclusively with the manufacturing of hardware and only using software in peripheral ways to achieve that main objective, Bosch has changed.
The company has been developing software of its own for quite some time and its Bosch Software Innovations business unit is likely to expand along with its industrial IoT business, which Hartung says is on course for exponential growth.
The main driver of the growth seems to be the demand for more information about what’s going to happen in the future. “The old way was a backward-looking analysis, and that was what was used to improve processes,” says Hartung.
“In the Bosch manufacturing system and some of the lean manufacturing systems, you now have this six-month forward-looking view. So you can forecast a scenario where you would like to be in six months with what we call system design of your factory, and you analyse data much faster – maybe half an hour or maybe an hour or two, depending on how you set up the system, after the data is collected and presented in the business information dashboard, and then go forward.
“With Industry 4.0 the vision is that you take data in realtime from the manufacturing field and use it for business optimisation and operational optimisation in realtime – and that’s obviously on a global scale.
“The connectivity is obviously just the base layer of this, so that’s the first thing you need to do – connect the data sources, which is the machines.
“You could just simply read the data on the controllers you have, and very often you see that – people just read out data from the controllers they have, which normally is data they would not have and don’t read out because normally these controllers are doing some mechanical job on the machine, controlling some axis, some movement, some assembly operations or something. The data they collect is just a read-out thing.
“But very often they use additional sensors, which then raise two questions. First, what data do you want to collect? And, second, do you now want to connect it through the internet to some system which is different to the manufacturing system – such as a data analysis system, which is often hosted in the cloud, or something which is different to the operational data collection system which uses the manufacturing controllers.
“And that is a connectivity system which we sometimes show as a gateway, which are small technical appliances which are located in the factory, near the sensors on the machines, which connect to the sensors or connect to the PLCs.”
“The connection can be Wi-Fi, they are in production, but in production sometimes you don’t want Wi-Fi, you want cables to be absolutely safe and for faster data transfer, but Wi-Fi is realtime pretty much.
“The data goes up to the analysis layer, and there you can decide how to implement improvements.
“The connectivity is the first step. That IoT controller is what we put on the old lathe.
“If you want to talk about speeds of data transfer, you have to distinguish realtime from time cycles.
“Ultra-fast realtime is the controller controlling a motion. Then medium-speed realtime would be what we would consider the IoT. And then the slower data flow for the backwards-looking analysis.”
Industrial dualism
In the past, factories were isolated places, kept well away from town and city centres, on industrial estates, as well as the internet.
Now the machines that operate in those factories are increasingly being connected to sensors, which collect data which can be viewed on a linked human-machine interface.
That still keeps the data within the factory, and it might be what some factory managers prefer.
The alternative is to connect the machines to an “internet gateway” device, which can be connected to either the factory’s on-site computers or the cloud through a virtual private network.
This creates something of a conundrum for those who want free-flowing data and yet would like everything to be secure and private.
Given that industrial espionage is as old as industry itself, it’s no surprise that the market is somewhat divided, as Hartung explains.
“There’s two philosophies,” says Hartung. “One philosophy is the group that says, ‘I need to definitely know what I need the data for, and I will deploy only the sensors for things which I know I want to use them for’.
“And with those customers, obviously you then go into the deeper into the technical things, and ask what is important, what is not important, where we should deploy them, what kinds of sensors, and other aspects of how it should be.
“We obviously have our consultants who would go through this process with the clients at our company. We also do external consulting for some companies.
“And then there is a second philosophy in Industry 4.0 where you say, ‘Wait a second, we don’t know yet what is going to become important, so let’s just deploy some sensors which take a whole bunch of data, in which some of them we don’t even know if we need them, but let’s see’.
“This sounds like an erratic approach, a non-technical approach, but it is actually a technical approach because in the modern sensor world, with these small MEMS sensors, for example, they are dirt-cheap.
“So it’s relatively inexpensive to deploy a massive sensor array with each one being so small and measuring nine, 10, 12 things simultaneously, always, and then later do the correlation analysis and data crunching, depending on what data you need at that time.
“So that second philosophy usually means you don’t need to do a lot of consulting and planning. You decide to go in first, go for data collection, data lake, data analysis, and improvement as a philosophy which says, ‘If I take too data much, it’s ok, if it’s over-heavy on the sensors, it doesn’t matter, I can scale back later – I can learn later’.
“Having said that, it’s not possible to do this in manufacturing without any context. If you don’t know what you’re doing, don’t deploy any sensors, right? But if you know what you’re doing, you can go with this overshooting approach, or scatter-gun approach, on the sensors – it does make sense sometimes.
“But obviously at some point later, you will need to have some intelligence about what data is being collected and what information you can take from that to improve your operations. It just depends on what you want to do.”
Sticking to the escrypt
Data privacy and security are big issues at the moment, and many companies have been struggling to stop hackers from accessing their systems.
Obviously if you have a machine which is not connected to the internet at all, there’s no risk attached. But such situations are increasingly a rarity.
Machines in their tens of thousands are being hooked up to the internet and the risk of hacking is obviously growing along with it.
Bosch’s answer is its own proprietary encryption technology and knowledge, which the company developed partly through buying a company called Escrypt, a data encryption specialist.
But internet security remains a tricky issue for most companies, says Hartung.
“The thing about gateways is that obviously you are now connecting your factory to the internet in some sense, and in whatever sense you do that, you have to be extremely sensitive that you protect your data and your factory,” says Hartung.
“Normally a factory is highly protected or even disconnected from the internet and now, with IoT and Industry 4.0, there is a connection necessary to some systems which are on the premises, your own cloud, which you have in the factory, or you could have a controlled gateway if you need a massive cloud, which you will not have in the factory because it’s not cost-effective or sensible.
“For the remote cloud, you need a controlled and secure data path and you need to know exactly what data is flowing out and what’s coming back.
“In most factories, you have a whole bunch of legacy systems, which is normal, and it will never be that the whole factory has PCs, PLCs, which are the most up to date.
“Therefore you have to be extremely sensitive if you do this gateway work, and make sure that these gateways are perfectly safe in terms of using encrypted handshakes and other key infrastructure technologies.
“That’s what we do. We use our Escrypt capability, from our own company, which operates in the data security business – we use their key infrastructures to encrypt and handle data.
“The owner of the factory wants the data to be transparent and wants to know exactly what data flows through the virtual private network. Otherwise, if you make a VPN protected channel from the sensor out to some other cloud, whatever it is, and there is no transparency about what is going through, it could leave an attack channel somewhere in the backhaul.
“So you have to be careful.”
Chips and everything
This interview was conducted in early May, 2017, a few weeks before Bosch’s announcement that it plans to build a billion-dollar semiconductor manufacturing facility.
This may explain some of Hartung’s reticence in answering specific questions about microcontrollers and microprocessors.
It’s unlikely that Bosch will build all-purpose controllers and processors. It tends to manufacture precisely what it needs, in that each of the chips it makes has a specific function in one or more of its technologies.
But Hartung seemed to suggest that the company could expand into manufacturing a broader range of processors and controllers.
“We have our own chip capability,” says Hartung. “We do a lot of electronics. Our fabrication plant in Reutlingen is capable of producing Asic [application-specific integrated circuit] chips and automotive electronics chips.
“But we don’t do standard microcontrollers and microprocessors.”
When it needs standard microcontrollers and microprocessors, Hartung says the company sources them from a variety of suppliers, including “Texas Instruments, STMicroelectronics, Intel and other large manufacturers, NXP… we source from everybody – ARM controllers… depending on what you need it for… but these controllers we source… Qualcomm is also a good source”.
He brings the conversation back to the Bosch IoT gateway and the components the company produces. “We make our own controllers. We have our own controller portfolio through Bosch Rexroth, and these are our own gateways,” Hartung says.
“Typically in these IoT devices you always need some kind of wireless connectivity, or at least most of the time, so you have these kind of chips from these kind of vendors in there, but normally you also have small controllers even on the edge device, which is the critical low-power device attached to the machine, collecting the data.
“There, normally you have ARM Cortex controllers in there depending on the capability you need. And then when you go to the internet gateway, you need something more powerful. It can then go up to desktop computers and other, more powerful computing machines.”
The crude oil of the 21st century
It’s been said that data is the crude oil of the 21st century, perhaps because cars and many other machines are moving away from petrol and going electric.
For electric cars to work properly, they need to store and process huge amounts of data. And if cars contain advanced driver assistance systems, which they increasingly do nowadays, they’ll need to process even larger amounts of data.
And in the future, when they’re fully electric and fully autonomous, they’ll need to process so much data that very few companies have the necessary computing infrastructure to deal with it. Not all of it can be done in the car itself.
Which may explain why BMW is building a data centre 10 times the size of its current data centre, and Volkswagen has become a customer of D-Wave, which sells quantum computers that cost $15 million each.
Large-scale, or even hyperscale computing infrastructure is fast-becoming an essential requirement for big manufacturers.
“Bosch has its own data centre,” says Hartung. “Obviously we need that, but we also use clients’ data capabilities because sometimes people want to use their own computers and servers.
“For our own production, our own data processing, we use our own infrastructure, also on the IT side to run optimisations.
“At the same time, if you want to be super-large scale, and you need massive cloud capabilities for a certain amount of time only, then you use an external cloud.”
Manufacturing customers differ, he says, with each having particular operational set-up and requirements.
“It’s what we have to discuss with the clients,” says Hartung. “Some clients are absolutely rigid in their policy, and say, ‘I don’t want to give anybody any data – just give me the algorithms and I’ll run it on my systems’.
“Which is fine, you know, if you want that.
“Others say, ‘I would like to use the cloud, if it’s safe and you guarantee me there’s a protection scheme around it’, and we sign the contract, and it’s ok.
“Normally our customers are conservative people, they like to keep things on premise. On the other hand, they know that some capabilities you just cannot build on premise – that’s not sensible because it’s just too expensive.
“I think that you have to see it’s all in development. This is a trend. The industry, with its thousands of machines which are mostly brownfield, old machines – some are new, some very few factories are built in this completely connected way – but the world is moving in this direction.
“There is not yet a common scheme, where people can say, ‘I want it this way or that way’. They’re all learning through considering their own specific business need for connectivity and data.
“I think this is very positive, that we’re seeing a learning spirit now coming to industry, especially in data analysis. This is, I think, the best thing to have come under the umbrella of Industry 4.0 – people are thinking and saying, ‘Is the way I did things for the last 20 years the way I would like to do it for the next 20 years? Maybe not, so let’s try out some new things’.”
Edge of today
Even in manufacturing, computing architecture – both hardware and software – is the critical factor in operations, and much of what Bosch does depends on getting that part right.
The IoT is easy to say, and perhaps easier to understand now that’s been talked about for a while, but it’s still difficult to create because of various factors, not least of which is lack of computing power at what is often referred to as the “edge”.
Hartung explains: “So you have three levels. The first level of the internet of things is what some people call the narrow bandwidth IoT, or some call the edge device IoT, and these are the small devices which have only batteries sometimes, are wireless, and have some limited functionality.
“Actually in these devices, programming-wise, you are in a complete crisis because you’re always in an energy crisis. So, as a programmer, you cannot go and program heavily with big libraries and so on, you have to code like 15 years ago, and be efficient, and even then you can only do very limited computation there.
“But the second level is the gateway level. There, you can do something. At that level, you have computational power available, but it has one ARM Cortex or whatever, so even there it is limited.
“Then you go to another, third level, which is maybe the on-premise cloud, which may have whatever – 100 servers or so. There you can do quite a lot of things, but it’s still incomparable to a remote-location cloud which might have 10,000 servers or 100,000 servers which you can unlock.
“So the new thinking would be, ‘Where do I do job? What do I do on-premise, what do I do on-chip?’
“You can filter some data in the sensor. You can say, ‘This is data I don’t send out because sending means energy consumption’. Every time you send something out, you drain some energy. So you have to decide in the sensor, ‘Do I send or don’t I send?’
“And then in the gateway, you can do a whole bunch of things and combine data, provide context, add layers to it, where the data is from, what machine it’s from, where the sensors are located, what’s the time code, and so on, and send it all up to the on-premise cloud.
“There you can do some heavy-duty stuff, data analysis, whatever is necessary, pattern recognition, deep learning algorithms can be applied, and so on.
“But if you want to do real heavy stuff, some monstrous data amounts, then you go off and go to the master cloud, which some companies will be capable of developing.”
A billion here, a billion there…
What Hartung didn’t say but probably knew at the time – and might have been referring to – was that BMW is building its giant data centre.
He also may have known that Apple has become a massive client for Bosch, which has been contracted to supply three sensors for the next iPhone, though he didn’t mention this.
But even though Bosch is at the forefront of so many technologies driving the modern world – literally in some instances – Hartung says the company is still learning, especially when it comes to the industrial internet.
“The challenge is always a technical challenge, it’s a learning challenge,” he says. “You have to discuss with people, approach people. You are dealing with legacy environments, which means you have always deal with uncertainty. Often you don’t know what’s going to happen.
“But to be honest I see it first as a giant opportunity.
“For example, we are doing machine tool equipment, hydraulic equipment controllers. If we were to add data capabilities to this kind of equipment, you are automatically into this multi-layer approach of data analysis and learning, and that’s a multi-billion-dollar business.
“By our own estimation, we will pass a billion dollars in the industrial internet market very fast.
“But there’s a second effect for us. We have hundreds of factories of our own, which means huge efficiencies through deploying these technologies, which is what we are doing now.”
Bosch has approximately 270 factories dotted around the world. “This also gives us the capability – which I think is differentiating – of inviting customers to come and see of the things we do.
“This automotive production may be on the periphery of what you want to do, maybe this is not your product, maybe this is not your process, but it’s worth having a look.”
Going, going… fixed
As many readers will know by now, the main feature of IoT in the industrial space is the ability to have a global view of your operation as well as know what’s happening to each individual machine.
This brings numerous specific benefits, as Hartung explains. “The main benefit is uptime optimisation, so machines don’t break, or before they even start malfunctioning, you see it upfront. That has a huge effect. Prevention is better than cure.
“But the second benefit is also clearly important: It’s a completely different level of productivity you can gain because you are much nearer to what happens in the actual processes.
“So you can see things really early, and transfer knowledge from different plants. If your plant in Africa has a system which is working better, why not upload it to the factory here in Europe? Or if it has a machine in China which is operating better than a similar machine elsewhere, why not take the system from China to wherever it is that will benefit from that data.
“These are huge advantages that can be gained. It might seem a dream to be able to do this across a global organisation, but it’s just a question of timescale – it just takes longer.”
“Through the maintenance and productivity analysis, you see very fast that it has huge effects.
“The potential is there, and it’s very interesting to see people doing these kind of things, when you see they’re sceptical at first, and then their eyes open to the possibilities of what they can do – which is often very different and more powerful that what they could do before.
“It’s not a miracle thing, but high productivity gains are possible. It’s not linear, it’s exponential.
“We are at the early phase of Industry 4.0, and Bosch also. We are not at the end of the curve, this is the first phase of the learning curve.”