The path of innovation is full of unexpected revelations. As you try different things and bring about new processes, you’ll run into things that surprise you. This was exactly what happened when we started to digitise our own production line at Fujitsu.
We were looking to apply disruptive technologies like Artificial Intelligence (AI) to our shop floor. But rather than achieving the quick results we were expecting through analytics, we found ourselves spending the vast majority of our time and resources on digitalisation.
This made me think about how we can better standardise the inputs we receive from connected machines on the shop floor, or the ‘operational technology’ (OT) so that they can interact with technologies like AI and machine learning within an edge cloud system.
It’s an unexpected area of interest, but it’s something that I see as incredibly important for us at Fujitsu – and for our partners and customers across the world.
So, in this blog post, I’m going to explain a bit more about what I mean by standardisation, and why it matters.
Uncovering a new problem
Manufacturers today know they need to transform to survive. And this means driving efficiencies on the shop floor by using analytics, big data, machine learning or AI.
However, most of these technologies depend on gathering data from the shop floor. Fully digitalising the shop floor has, therefore, become a prerequisite for exploiting pretty much every disruptive technology. This is tricky for the vast majority of manufacturers, who don’t have the know-how required to manage these technologies in the same ecosystem as their operational technology – which until now has been running as an isolated entity.
Feedback from our own manufacturing lines, as well as those of our customers, has shown that the digitalisation process often does not go as desired. Instead of spending the larger part of our time investigating the insights from the data, we have found that most of the cost and time investments end up being spent digitalising the shop floor.
On top of this, most of the implemented systems are hugely costly to maintain. All it takes is a simple sensor exchange somewhere with a new (and of course incompatible) model, or some machinery firmware upgraded in an inconsistent way, and suddenly you have to go to huge lengths to keep the data models workable.
This architecture endangers every application of new technologies. And it’s exactly this problem that standardisation can solve.
Simplify everything with just one language
You have machinery installed along the production line. These machines are already able to provide you with inputs – but they provide them in their own scale. One sensor, for instance, might give you the temperature in Kelvin; another might use Celsius; another might provide you with information on its very own scale.
This is incredibly confusing and difficult. An enormous amount of time is spent converting all the different scales into a single one.
What would make this process more efficient? A standardised, common language that would save you from having to deal with the translations of digital languages.
OPC UA turns out to be such a standard. It converts every machine and every sensor into a standard model. This includes status variables and events as well as methods to access the machine. Variables get valid value ranges, and units are suddenly much easier to control.
Originally the standard was designed as a means for machines to communicate with each other more easily, but like all good innovations, it had a surprising secondary application! Since it’s a communication standard capable of preserving and transporting the semantic data on an application level, we can use it to talk to other software instances located somewhere in the cloud – like AI software.
Working with a community to solve industry problems
At the moment, it’s still too early to exactly predict the impact of this new way to standardise communication. In my view, it’s definitely worth a thorough investigation – and that’s exactly what is being done as part of the DIN SPEC 92222initiative.
I recently attended a working group meeting of this initiative. Their goal is to create some sort of agile standard, which essentially means consolidating their current understanding into a working document that over time could eventually become a real standard.
The group is made up of manufacturers who have been in the business for decades, and what they share is down to earth and very relevant. I found their insights extremely valuable.
Making manufacturing work with the cloud
One of the topics we covered in the working group involved the ecosystem of developing a communication standard that allows OT to communicate via the edge with multiple clouds, and also facilitates cloud to cloud communication.
Most manufacturers will need a multi-cloud strategy. There will be situations where a manufacturer has their own applications, developed on a selected Internet of Things (IoT) platform belonging to a cloud vendor. But this doesn’t mean that all communication will run just between the OT/edge and this cloud.
Most manufacturers have machinery in operation that belongs to multiple different vendors. A growing number of these vendors will offer their own digital services like predictive maintenance, or pay as you use consumption models. And they will provide it from their own clouds (even if they’re hosted by one of the prominent cloud infrastructure providers, they’ll still appear as separate cloud architectures).
This represents a complete shift away from the ideas of device operation that we work with now. Making it work will require a manufacturer’s IT infrastructure to maintain relationships with multiple clouds from different vendors, and their own cloud application services.
Standardisation: the answer we didn’t expect to find
The full complexity of the manufacturing digitalisation process has come to light. In an ecosystem with so many moving parts, there was bound to be some tension when it came to making it all correspond.
This is why standardisation could be such an important avenue of future inquiry. To derive the full benefit of AI, big data, and predictive maintenance – what I have often referred to as ‘disruptive technologies’ – we need reliable inputs from the shop floor that are easy to use.
It’s important to remember that digitalisation itself doesn’t offer manufacturers the benefit. Instead, digitalisation is a necessary step you must undertake in order to access the real benefit, which comes from applying the disruptive technologies to the shop floor.
Standardisation could prove to be the catalyst that enables us to use digitalisation to extract the efficiencies we are looking for. I say ‘could’ because we can’t yet be certain.
There’s still more work to be done in this area – and I’m looking forward to finding out what we discover.