Data science in pictorial form Add to my magazine Remove from my magazine

Applying Data Science to ABB turbochargers

With product lifecycles that can easily last for decades, ABB’s turbochargers are tested extensively to ensure longevity. We’re not just talking about future products and R&D, however, and ABB Turbocharging is constantly looking for new ways to improve and ensure our existing products are running as efficiently as possible. Data science is just one of the ways we’re using intelligence to further enhance our digital customer solutions to help understand how turbochargers work both in testing and the real world. Gianluca Rondinini, Data Scientist at ABB Turbocharging, tells us more.

Data science is the practice of applying advanced analytics to extract valuable information from data. It’s a field that’s seen huge growth over the past 10 years across different industries, and with a wealth of data at our fingertips, it’s particularly relevant for ABB Turbocharging and our customers.

Building powerful analytics for our external and internal customers

Gianluca joined ABB Turbocharging three years ago and works closely together with a team of data engineers and software developers. “We build next-level analytics applications,” Gianluca explains, “including web-based applications to fuel our Digital Customer Solutions. We show and plot data that enables our external customers and internal users such as technical service teams and the technology R&D department to generate analytical insights to drive business decisions.”

Data is captured from several different sources, with plenty of sensors on modern engines already monitoring different operating conditions. Useful data points could be something as simple as the variance in temperature of a particular component, and embracing data science can make sure that anything out of the ordinary won’t go unnoticed.

“Anomaly detection is probably our main area of focus at the moment,” Gianluca explains. “We could be looking in the data for bigger issues, such as the turbocharger potentially breaking down, or perhaps not running as efficiently as it could, but we also look for smaller problems, such as sensors that may be working incorrectly.

"We evaluate the data and generate insights which can help people from other departments to take action, and this kind of insight can help at any level"
Gianluca Rondinini

Understanding equipment behaviors starts with identifying and filtering anomalies

“The turbocharger may appear to work perfectly fine, but our analytics will pick up on data where sensors are reporting crazy or unusual numbers, helping us to highlight potential issues. That’s the sort of thing we detect. With such use cases, we’re also dealing with deep learning, and the more data we’re able to gather, the more we’ll be able to do in the future.

”For our internal business, we evaluate the data and generate insights which can help people from other departments to take action, and this kind of insight can help at any level.”

“It can help when it comes to servicing, for example, analyzing data from turbochargers which have been up and running for a long time, and it can also make a difference when it comes to designing new products, when it comes to understanding the kind of stresses a turbocharger sees in the real world. And using that data and the knowledge we gain from analyzing it, we can shape the product accordingly in terms of physical specs, in a way that perhaps lasts longer and is less likely to see issues under load.”

Making our customers’ lives easier through powerful analytics

Developing the new analytics applications has proved an exciting challenge in itself, says Gianluca. “Machine learning models have usually been developed in completely different frameworks, and for completely different purposes, so we’re using models that weren’t originally developed for use with turbochargers. Converting and adapting these models has really brought the best out of the team. It’s not like you can just search the internet for information about using such technologies for industrial purposes; the literature is quite shallow and there’s a lot of work required to adapt such things.”

As more information is amassed and baselines are created, Gianluca and his colleagues are also looking at the possibility of analyzing data in real-time, with the team looking towards artificial intelligence to enhance their digital models. It’s a technology that compliments data science perfectly, says Gianluca.

“AI and data science definitely go hand in hand. There’s so much potential to use deep learning within data science, and both fields are ultimately looking to solve real-world problems using data.

“There’s a lot of value to the data we’re gathering, but that’s really just the beginning. To make an even bigger difference, and to help generate tangible results, you still need to actually extract value from that data, and ultimately that’s the role of the data scientist. That’s where we can work with other departments and customers, and really add value from the data we’re gathering.”

“I believe the technology will shift more toward research, and with help from deep learning and AI, the information and analytics we’re generating will become a lot easier to deal with. That will make a big difference to both internal and external customers, and ultimately, our goal is to make their lives easier.”

Image credits: sdecoret/Shutterstock