Dr David Bott, Principal Fellow - WMG

Recently I was invited to a roundtable discussion on the future of the smart factory. It was couched in terms of Industry 4.0 and majored on the disruptive nature of the changes facing companies in the manufacturing sector. I was asked to lead the discussion by talking about my experiences.

Throughout my career, I have observed that people like to talk about disruption and step-change, but I'm not sure they are as good as predicting them as they like to think. Big changes do happen but, in my experience, we only recognise them with hindsight. And often step change is code for a steep learning curve and a reason not to act. The main reason for predicting them I have seen seems to be that it tends to unlock large amounts of funding from people scared of the change!

Manufacturing is the word used to describe how we turn raw materials into products on a large scale. Given that most products are complex assemblies of components and sub-assemblies, most start-to-finish manufacturing processes comprise many companies in a supply chain, but it is, for the most part, a very physical activity.

Over the last couple of decades, with the increasing power and ubiquity of computers, we have created a new industry which performs a similar task by transforming raw data into useful information. We have rigorously applied it to several commercial activities already and used it to change business models, often with new, insurgent companies being more successful at the change than pre-existing companies. That it would be applied to manufacturing was really never in doubt, but I would assert its been going on for some time already.

I started in the chemicals industry a number of decades ago. At the beginning of my career, most plants that I visited were staffed by people who turned valves and used recipes that had worked in the past. When things went wrong, they relied on experience to tell them what to do. But there were a few people trying to model these plants using their understanding of fundamental chemistry, they could predict how a plant would be affected by changing inputs or control parameters. These models ran in batch mode, overnight on computers with paper tape inputs and a semi-religious staff. As the speed and capacity of computers increased, these models could be applied in real-time, and then faster than real-time, enabling predictive exploration of control options. By the time I left the chemicals industry (15 years ago), most plants were controlled by a very small group of people who effectively played a computer game which then controlled the plant.

As the power of computers increased even more, they could model activities outside the pipework. Stock control, the input of raw materials, the timetabling of deliveries to customers were all included in a business that effectively ran inside a computer and what was efficient was imposed on the real world.

Making undifferentiated products is one thing, but the combination of what has happened in the retail industry and the desire for mass customisation of a wide range of products, there is now the imperative to transform the manufacturing parts of many supply chains.

Nowadays, everything can be monitored, controlled or logged digitally. It starts with product specification. Whether we use drop down menus from a website, or an algorithm that interprets what we say, then what we want is increasingly stored and communicated as a digital file. Once this specification is checked against manufacturing capability, the seller forwards all or part of the order to the factories that make/assemble the necessary components or products. This can be used to schedule raw material orders, control manufacturing machinery. Logistics are almost completely digital these days, with customers able to see where their delivery is and when it is due.

One thing that is new is all that information that now has to move up and down the supply chain. The first question is who owns it? There were stories of those near the end of the supply chain using the detailed information they could obtain and analyse as a way to control pricing. There were fears that machines, or even whole factories, could be hacked either to disrupt production, or extort money not to.

What all this means is that what goes on in factories is changing. The core skills or understanding about how a material is transformed into a component will increasingly reside with those who design and build factories, not necessarily in those who run them. Once they are up and running, these factories will need only management and maintenance. It is very likely that both of these functions will eventually be automated we can now monitor and implement preventative maintenance regimes and machinery is often designed with robot maintenance in mind. Like the chemical plants I described earlier, the factory and its associated supply chain will be run by someone effectively playing a computer game that decides the specifics and tells other machines what to do.

And therein lies a challenge for us all. When you design a self-managing factory (be it large or small) you need to profoundly understand what you are trying to do, anticipate variances from normal behaviour and know how to address them. If you miss out on the response to a circumstance that does arise and the control system doesn’t know what to do, then problems occur. Over time machines learning systems could be introduced, but they have to learn by making mistakes, so complete and thorough programming at the beginning will always be the preferred option. This is a version of the argument going on at the moment about driverless cars, where it induces quite a debate but in a wider audience.

What transpired in the conversation over dinner was that many SMEs are put off addressing the issue in a strategic manner because they have been persuaded by what they read that it will be a disruptive and difficult change, although some of them have already made the sort of changes I saw in the chemical industry. Perhaps we need to recognise that the hyperbole needed to get funding agencies to hand over the cash can get in the way of working with the companies that will translate all these new ideas into commercial success?

For me, this is the point. In the end, all this digitally enabled manufacturing has to result in physical products. Unless the process we are implementing delivers the goods as intended, then it will all amount to nothing. Whether this is truly a step-change or just a steep learning curve will only be certain with hindsight, but given the radical changes digital technologies have wrought in other parts of commerce, we would be crazy not to design and build a path to our own digitally enabled factories.