My name is Jeroen Didden and I started my PhD in May 2020 within the Operation, Planning, Accounting and Control research group of the Department of Industrial Engineering and Innovation Sciences. I’m doing research on the transition to Industry 4.0. Companies are collecting all sorts of data, but they’re not really using this to enhance their systems. Our goal is to help them make better use of data to improve certain key performance indexes within factories, mostly regarding scheduling and planning systems.
I started in the ‘The Digital Factory of the Future’ project group, an initiative by Brainport Industry Campus, with the idea to create a digital blueprint for future autonomous factories in which no humans are involved. Multi-agent systems are often proposed in literature; these are entities (like robots or finance systems) that transfer information between one another to enhance operations. But what I noticed is that trying to implement these all at once is a big step. My idea was then: how can we help companies to understand that transition?
Instead of creating very advanced algorithms immediately, how can we take an intermediate step so that they understand what the algorithm is doing? This gives them an idea of what Industry 4.0 can mean for them. Within our group, we have a lot of PhD students in semiconductor or high-mix-low-volume manufacturing, which gives us many opportunities to use the knowledge of others to improve our project. If a student has come up with a good algorithm, we might try to use it in someone else’s case to see if it also works there or to expand on it. If they do machine scheduling, for example, we can add autonomous vehicle scheduling or vice versa. Sometimes, we try to combine various projects because we then get a sense of the complete factory instead of just one small piece. Collaboration is what puts it all together.
The challenge from our perspective, of course, is creating the algorithms. But from an industrial perspective, the challenge is often that companies are used to scheduling by hand, so they know their products and when to plan them based on experience. Usually, companies do their scheduling by choosing the product that has to be done the earliest. We’re trying to introduce more complex rules so that we can look at combinations such as the products that need to be done the earliest, have the shortest lead time and are of a certain job type. We then combine all these rules into a single rule and use this for scheduling.
Compared to standard methods, our first results show a very good improvement in terms of how close to a due date we can finish a job. In the new environments we’re considering, the minimum improvement is already around 20-30%. Through this, I want to make the transition to Industry 4.0 easier for factories. What I find very interesting is how autonomous factories might work: all those vehicles driving around, carrying certain products, delivering them to machines that pick them up automatically and process them. And that’s also something that companies themselves desire, so I hope that I can help take an additional step towards this.