Learn how the pharma industry is catching up on Industry 4.0 and watch a film about a MADE research project on implementing Industry 4.0 at FUJIFILM Diosynth Biotechnologies, Hillerød, Denmark.
The general consensus is that the pharma industry has been a slow starter in factory digitalisation through digital twins, IoT, Big Data, AI and machine learning – otherwise known as Industry 4.0 (I4.0) technology.
This is due to stringent regulatory requirements and intrinsic caution concerning consumer health.
This means the unrealised gains from I4.0 are considerable. According to a survey conducted in 2019 by the US consultancy firm Bain & Company, “Pharma executives expect smart connected factories [ed. I4.0) to produce total savings of 20% or more, while improving quality and making deliveries more reliable.”
This is currently being explored. Now machine learning and AI are increasingly used in drug discovery and the technologies are also finding their way into the manufacturing processes:
“New modular sites designed in the last three to five years, and which are going into operation now, often embrace Pharma 4.0 concepts at many levels,” Wolfgang Winter, Agilent Technologies to The International Society for Pharmaceutical Engineering (ISPE) in an interview, who adds that older sites probably will stick to “only very focused, gradual investments and upgrades”.
Digitalising a pharma factory, Hillerød
At the newly opened FUJIFILM Diosynth Biotechnologies (FDB) factory in Hillerød, Denmark, Breno Strüssmann from the Technical University of Denmark (DTU) has made it the mission of his MADE PhD to help the company enter the Industry 4.0 era.
“The goal is to raise awareness of the possibilities of Industry 4.0 technologies and help implement them,” he explains.
In the project, Strüssmann has dived into three areas:
- The application of machine learning to the chemical process of drug manufacturing in order to increase yield.
- In collaboration with FORCE Technology, a proof of concept demonstrating how quality control of injection pens could be further digitalised through an IoT solution connecting the quality control machines to cloud data which is instantly sent to the operator via an easily read digital dashboard. Previously, the operator had to walk to the machine to read the results.
- Measurement of the company’s digital maturity level. Through 20 stakeholder interviews, he has created a baseline for the current digitalisation level and an overview of barriers to reaching a higher level.
MADE research project
This research project was carried out by:
FUJIFILM Diosynth Biotechnologies (FDB), DTU and FORCE Technology
The project is part of the MADE FAST research platform focusing on Flexible, Agile, and Sustainable manufacturing enabled by talented employees.
MADE FAST is co-financed by participating companies and partners in the manufacturing ecosystem. The main sponsor is The Danish Industry Foundation.
Increasing yield through machine learning
Zooming in on the machine learning project, Strüssmann is especially excited:
“Machine learning requires a lot of data, so we can still improve, but the positive results of testing show that we can increase the final yield,” he explains.
The final amount or yield of medicine produced depends on the efficiency of manufacturing processes. This can be divided into upstream and downstream activities, where the aim for the upstream process is faster production of a large mass, which is then handled in a downstream process in which medicine from that mass is separated and purified.
Focusing on the upstream process, Strüssmann is using machine learning to determine the right parameters (pH-value, temperature, etc.) for optimising cell growth
“The upstream process begins with cells in a small flask and by the end we have cells in 20,000-litre bioreactors. It’s very important to measure all the parameters throughout the whole process because we don’t want to lose such a huge batch – right?” Strüssmann explains, adding:
“If the batch has the wrong pH-value, say, the cells start to die for instance. We want to reduce the number of lost batches to zero.”
Machine learning
Machine learning (ML) is a type of artificial intelligence that allows machines to learn from data without being explicitly programmed. It does this by optimizing model parameters (i.e. internal variables) through calculations, such that the model’s behavior reflects the data or experience. The learning algorithm then continuously updates the parameter values as learning progresses, enabling the ML model to learn and make predictions or decisions based on data science.
The applications of machine learning are wide-ranging, spanning industries such as healthcare, finance, marketing, transportation, and more.
A hybrid between ML and mathematical models
However, cell growth is not as predictable as one might have hoped.
“When you’re dealing with live cells, it’s hard to predict how they are going to form and evolve in time because there are a lot of parameters and they all have different, dependencies amongst each other,” Strüssmann explains and continues:
“What’s, coming up in recent years is that you apply hybrid machine learning where you have both mathematical models – which we already know – and historical data to predict how your batch is going to be. This is coming in the near future.”
This is coming in the near future.
Breno Renaut Strüssmann, MADE PhD
This has also been the approach in the PhD project, where the historical data is based on previous batches.
What next?
The results from the three areas researched by Strüssmann will be passed on to FDB. The IoT solution for quality control is currently being installed and Strüssmann will continue to work with FDB in the position of Digital Twin Lead in the Global Technology Development team.
In this role, he will be gathering data to further improve his machine learning model as well as continuing to help FDB benefit from Industry 4.0 technologies in general.