Sheila Whelan - Using Data to Optimize Manufacturing

“Advanced data analytics can interrogate data at a much greater level than people realise, and can find interactions and patterns that the human brain is incapable of deciphering, and that can have direct value.”

For Sheila Whelan, a specialist in data analytics and machine learning with DMI, a career in technology wasn’t initially on the cards when she started her working life. Her undergraduate degree from the University of Limerick was in industrial chemistry but it was her first interaction with ‘regression modelling’ in an early job that made her rethink her trajectory.

“I started out working as a process engineer at first, a kind of chemical engineer, and that brought me into contact with lots of different people doing lots of different things. In particular, I was working on a large site with a lot of automated data collection from instrumentation and equipment, and a colleague generously involved me in his work on regression models and that just blew me away,” she said.

“Regression models are part of the basis of machine learning, and they use historical data to create predictions. I learned how analysing already collected data could allow us to predict the next week or two of activity. It was really powerful to see what could be done and figuring out which parameters actually impacted the accuracy of predictions most. I was hooked.”

What followed was a new enthusiasm for what could be done using data and the insights that could be derived from it. Before working on these regression models, Whelan had been only vaguely aware of data science but the experience led to her returning to college to study data analytics and take her career in a new direction.

“I want to be really interested in what I’m working on a day to day basis, and I have half of my career hopefully in front of me. So it’s important to be willing to make the changes that I need to stay engaged. I want to be able to lose myself in being deeply engaged with my work and I’m lucky to be able to say that’s what I now do,” she said.

Today Whelan is an optimisation data scientist with DMI, where her job focuses on helping customers understand their data flow and identify areas where data can contribute to growing efficiency, reducing cost and increasing capacity.

“It’s a great time to be working in this area because there is growing awareness that data analytics is about a lot more than just trends over time plotted on a graph. Data is becoming core to how lots of manufacturers think about what they do, and those that are good at what they do are pulling away from those that aren’t, often because of this difference,” she said.

“Advanced data analytics can interrogate data at a much greater level than people realise, and can find interactions and patterns that the human brain is incapable of deciphering, and that can have direct value.”

Whelan works from a position of having a deep understanding that everything that happens in a manufacturing environment is subservient to the business goals of the company that owns it. Each production environment is different and operates to a certain level of efficiency, at a certain cost, with a certain tolerance for flaws and wastage.

But there is always a gap between how an environment is planned and how it actually operates. For example, she says, it’s possible for a particular machine to generate more heat than was expected and for that to have an impact on the overall process.

“So if you’re analysing your data in a clever way, you can discover that and perhaps find that reducing that temperature by half a degree might result in an improvement in your quality output. But if you’re not looking, you won’t find that. You don’t know what you don’t know,” she said.

“And using machine learning, you can take that a step further where the system will automatically change set points further back in the process to correct the issue before it arises.”

Part of Whelan’s work in DMI is to not just be at the cutting edge of what’s possible with data analytics but also to be an advocate for customers in helping them understand what can be done specifically for them. Often with large organisations, funding decisions are made at senior management or board level by people who can be quite far removed from the technical challenges that are being addressed.

But nobody wants to spend money without knowing exactly what they’re paying for and why. They want to know how they can assess whether a project will be a success or not, and to do that, they need to have potentially deeply technical subjects explained to them in a non-patronising way.

“The best way to get the benefits of working this way across is to show people an actual example of the technology in action. We might say to someone, with these seven process variables, you can predict this quality parameter to this level of accuracy,” said Whelan.

“We can also do things like take a customer’s historically collected data, and use three quarters of it to train a model but keep the other quarter to test the model. The idea is that the model has never seen that quarter of the data before, so you can prove how good your predictions are by running it on the test data. And when someone can see that in action, then they can make an informed choice.”

Looking ahead, Whelan thinks we’re on the cusp of a data revolution. Processing power and sensors are becoming cheaper, AI and machine learning are accelerating in terms of what they can deliver and at what price points, and generally, the future of manufacturing is data-shaped.

“I read recently a statistic that said that 80 per cent of all the data out there in the world was created in the last two years. That’s how exponential and explosive the growth of data is and there’s no reason to think that will stop,” said Whelan.