The Great Industry - Heurisko - take aways
Industry 4.0 marks the fourth industrial revolution. What does that mean for us? Are we part of it, i.e. is the industry ready for big data and A.I.? Time to find out! Dataroots was invited to Heurisko 2022, the annual seminar hosted by Flanders Make, where the most innovative and industry-ready research results and applications are presented.
Industry Zero to Industry Hero
There have been multiple industrial revolutions in the last 200 years. The first industrial revolution happened around the 18th century where the use of steampower and mechanisation of production was the biggest driver. The second industrial revolution began in the 19th century through the discovery of electricity and is known for the invention of assembly line production. The third industrial revolution is a bit more recent. It started around the '70s and consisted of partial automation using memory-programmable controls and computers. This leads us to the fourth industrial revolution, the one we are currently in. Its most notable characterization is the digitalisation of manufacturing.
Flanders Make aims to be the bridge between the industry and research, making Industry 4.0 as relevant as ever for them.
They conducted a survey in 2021 to determine what currently drives companies towards the future. Sustainability, digitalisation and big data seem to be the general trends, which is not unexpected.
Heurisko allowed us to gain insights on how Industry 4.0 and A.I. go hand in hand. Although A.I. was not the focus of this conference, it became clear during some of the talks that the industry is ready for A.I. But what exactly does that mean, being ready for A.I.?
In the following section I will discuss a few talks of Heurisko and where A.I. fits in.
Applications in the industry
The first application is about finding more advanced cooling methods for drivetrains, which is the group of components of a motor vehicle that deliver power to the drive wheels. Technically speaking, this is the domain of industrial engineering and quite far from big data and A.I.. So, how can we contribute to such tangible problems?
Testing out all proposed solutions for these problems, requires a lot of experiments. This is not only a costly process but also a very time-consuming one. One way to improve this process using A.I. would be to create models that can simulate and predict the behaviour of different cooling methods on drivetrains. This way, as a supporting tool, some of the physical experiments are not necessary because we have these simulations.
The second application I would like to talk about is less of a direct application but more of an implication. It comes from a presentation with the title ''Tackling the energy transition in manufacturing: an industrial DC grid''. The idea behind this is that we can migrate our AC grid back (yes all the way back to its roots), to a DC grid. The first experiments done by Flanders Make concluded that while it seems to be more challenging to generalize this than originally thought, it still deems useful for a lot of use cases. Since these use cases are still specific, they launched a platform where companies and individuals, can give certain parameters and they will receive a report on what benefits they would gain from changing to a DC grid.
The last application is an application with a lot of use for consumers. It is about how A.I. can help paint shops. Consider a paint shop that paints different kind of metals. If they get a job for which they need to paint 500 metal boxes, they would need to manually paint one and record the process as well as indicate points on the chair for the robot to recognize. The process of manually programming the robot in total can take up to 12 hours for only programming one face of the box. There are two favorable ways A.I. can help this company. One of them would be to use a digital twin. This would allow people to work from the comfort of their own living room. You can extend the environment using A.I. to mimic the way the paint sprays on a certain surface to allow for even smoother and quicker programming of the robot. The other way would be to implement computer vision in a way that you would no longer need to manually paint. The difficulty here lies within the surfaces, it is very hard to simulate the behaviour of the paint when there would be a bump in the surface.
As we can see, there is a lot of room for A.I. in the industry. Sadly all these applications are could be's, meaning that they are currently not applied. Why is that?
As mentioned in the beginning, one of the keypoints of Industry 4.0 is big data, but as we can conclude from the applications, it is currently still a far way out. Even though A.I. is being succesfully applied in a lot of domains, which are mostly consumer-centric, the industry still struggles a bit. I do not think it is because of a lack of knowledge, bad practices or even a lack of tools.
I feel that it is moreso the lack of data that is available on a large scale. This is due to IoT devices to collect data being costly in recent years. This resulted in data being harder to collect, especially on a large scale. For example, to be able to predict the savings of your DC grid as compared to your AC grid, you need a lot of historical data with the actual savings in order to predict the next.
Luckily, the methods to create synthetic data (simulated data) are advancing rapidly. Based on an article that appeared in Wall Street Journal, it is predicted that by 2024 more than 60% of the data used for the development of A.I. will be synthetic.
We are moving in the right direction, where we lack data, there will be synthetic data available to be able to simulate a production environment. This will allow us to apply A.I. in order to support research towards a better, more sustainable future.
Not only this though, it can also lead to the absolute personalisation of all things for the consumer. Just imagine, you take a pen and paper and draw the craziest thing you can imagine which you want in your living room. Because of the automation of manufacturing and its accompanying intelligence, it should be perfectly feasible to manufacture whatever you draw and paint it in the most crazy colour combinations you can imagine. Couple this with delivery at hyper speeds and you receive your item the next day. That would be the dream, right? We are still very far from that, or are we?