Personalisation will drive the next wave of manufacturing.
The European Union is keen to win back manufacturing but has, so far, limited ambitions. In 2012, the group of nations set a target for manufacturing to represent 20 per cent of ‘total value added’ by 2020. To get there, European manufacturing sector leader Germany needs do very little - it’s the other nations that are letting the side down.
But Germany is the spiritual home of the technological fightback. Having supplied many highly automated machine tools to Chinese factories, Germany has taken centre stage in the Industry 4.0, or Industrie 4.0, movement. And the country is spawning companies such as MyMuesli, which uses an automated plant to produce individually mixed boxes of breakfast cereal that employ the concepts behind Industry 4.0.
The country set up an extensive programme of R&D projects to develop technologies and protocols for more flexible industrial automation. The programme is possibly too extensive, according to Rahman Jamal, European technical and marketing director at National Instruments. “The reference architecture being developed seems to be covering too many things at once,” he comments.
Since the programme got under way, Industry 4.0 has expanded internationally, leading to the creation of groups such as the Industrial Internet Consortium (IIC), which has promoted the development of a number of multi-organisation testbeds to try to work through the issues the architecture raises. Not least is the question: what is Industry 4.0?
Cornelius Baur and Dominik Wee of management consultancy McKinsey wrote in 2015: “We define Industry 4.0 as the next phase in the digitisation of the manufacturing sector, driven by four disruptions: the astonishing rise in data volumes, computational power and connectivity, especially new low-power wide-area networks; the emergence of analytics and business-intelligence capabilities; new forms of human-machine interaction such as touch interfaces and augmented-reality systems; and improvements in transferring digital instructions to the physical world, such as advanced robotics and 3D printing.”
The emphasis on digitisation and the adoption of sensors and robotics looks at first sight as though it simply marks an evolution of Industry 3.0 - the automation revolution in manufacturing that began in the 1970s.
“It’s definitely not a revolution. It’s an evolutionary process,” says Jamal, “but it has definitely moved beyond the hype level. We get approached not only as a platform provider but also to make their ideas Industry 4.0 compatible. And we are now demonstrating proofs of concept.”
The key difference between Industry 4.0 and prior generations of industrial automation is the focus on real-time, large-scale data analysis. This is intended not just to tune production on the fly but to incorporate information about demand and distribution to get products into customers’ hands faster. Some of this is likely to displace traditional engineering practices.
One of the driving forces for Industry 4.0 is mass customisation made possible by changes in manufacturing techniques. Clothing and printing companies have not waited for the named revolution. In Shoreditch, London, startup Moo uses printing presses made by HP that dispense with the need to make physical plates. They form an image photographically that electrostatically attracts ink that is then pressed onto paper. The Indigo printers make it possible for Moo customers to upload not just custom designs for business and greetings cards but a different design for each card within a batch of 50 or 100. Scheduling software mixes and matches designs from different jobs to optimise throughput through the digital presses. They are then separated when the cards are diced and sorted.
In Germany, MyMuesli makes customised breakfast cereal for customers out of a collection of base grains, nuts and fruit. Professor Wolfgang Wahlster of the University of Saarland says: “Each package moves around the factory on an intelligent product carrier. The carrier tells the filling machines, according to the shopping list, ‘I want this and this’. It’s an inversion of normal manufacturing design.”
Having a product determine how it is to be made is not unique to Industry 4.0. Factories that call for high levels of customisation have implemented this approach for years. For example, more than five years ago power-supply manufacturer TDK-Lambda moved its plant in Ilfracombe, Devon, to a system where the digital product memory is conveyed as a barcode stuck to the PCB. This code controls what components each power supply receives before they are soldered into place and inspected.
TDK-Lambda also reworked how it inspects the power-supply boards to avoid delays in responding to problems during soldering. Traditional manufacturing techniques queue products for inspection. The inspection may find the soldering is not up to par because of problems with the chemicals or chamber temperature or that one of the pick-and-place machines is drifting out of alignment. If the product with the problem is at the head of a queue, it means all those that follow are likely to have similar defects, requiring a lot of rework. The company upgraded its automated optical inspection hardware to support real-time inspection to avoid the backlog building up.
A more widespread shift towards Industry 4.0 techniques will increase the use of real-time image processing, says Giles Peckham, EMEA marketing director at Xilinx: “There is a lot of overlap there.” This is likely to increase the use of programmable logic, as these devices can be used to act as coprocessors for conventional microprocessors. Both Altera, bought by Intel to augment its server processor business, and Xilinx make system-on-chip devices that combine programmable logic with standard processors.
Today’s customisable manufacturing processes generally revolve around a set of predefined options. Potentially, processes designed for Industry 4.0 could offer deeper levels of customisation.
At NI Week last year, National Instruments CEO James Truchard said it would be possible for users to come up with designs for their own bicycles and send them to a factory for manufacture. To avoid the custom bike shapes from collapsing as soon as the rider gets on, the manufacturer will need to constrain the plans and show designers what is possible and what is a bad idea. It means bringing together simulation and test results in a computer model to predict how the tweaked version is likely to behave.
“You will have to use your previous test data to predict how this new product will work, to ensure that this one-of-a-kind bicycle gets made correctly,” says Truchard.
Some of the test data is likely to come from ongoing condition monitoring. Jet engine makers GE and Rolls-Royce now routinely collect data from their products as they fly around the world to schedule maintenance. In doing so, they stand a better chance of reducing downtime for individual aircraft. It is this possibility that convinces airlines to pay the engine makers for their services.
Rob High, CTO of IBM’s Watson artificial intelligence unit says one of the key aims of his operation is to “understand the language of machinery” and use that to identify and avoid maintenance problems.
Even the product carrier may come to learn the language of the production machinery. Moving from barcodes to processors with RF links will reduce the need for machine tools to access and wait for instructions from remote servers. Messages about each product’s progress will still make their way back to the cloud servers that can be used to inform overall strategy and tell machines downstream to reschedule operations as different products turn up. But the product itself can influence its route through the factory.
“A product can tell equipment that it needs a green production cycle. But, if the product needs to be produced in a hurry for the customer, it can take a different route and the machinery will react to that,” Wahlster says.
For all that data you need standards and the use of a lingua franca for the different machines to communicate with each other - both inside and outside the factory. This is potentially the biggest sticking point for factory owners today.
Today’s factories rely on a hodgepodge of largely incompatible communications protocols at the machine-tool level. These grew out of the move to fieldbuses during the 1990s, representing industrial automation’s pre-Internet phase. Generally a low-throughput, real-time link between tools and the programmable logic controllers (PLCs) or industrial PCs that coordinate their activities, many of the fieldbus options are tied to specific vendors - often specialists in specific parts of the production process.
“There is a jungle of fieldbuses,” says Wahlster. “There can be as many as ten types in one factory.”
The answer is to do what office computing did in the 1990s - move to IP-based protocols. That decision is fairly uncontroversial. The more difficult question is which collection of IP-based protocols? The decision for the IIoT is perhaps easier than that for the wider IoT. A year ago, two of the main groups working on international industrial standards - the Open Interconnect Consortium (OIC) and the IIC - decided to work together. The IIC has organised the creation of a number of testbeds by companies and research groups to work out interoperability issues.
“We are involved in four testbeds,” Jamal says, including a time-sensitive networking version of Ethernet. “It’s physically hosted by NI. Once this testbed is up and running, by the SPS Drives show in November probably, it will be with the IIoT community.”
On top of these core protocols, companies operating in vertical markets are defining languages that machines can use to communicate. An example is OML, which is intended to let the machines used in electronics assembly talk to each other directly. But the key to progress for any of these will be agreement to use them at the human rather than the machine level.