Epiroc advances manufacturing innovation with Azure Machine Learning and Azure Data Factory

Improving innovation to deliver highest quality

Driven by innovation, the Swedish manufacturer and market leader Epiroc began producing tools for the construction of the Swedish railway network in 1873 and has evolved to become a global leader in supplying the mining and construction industries with advanced equipment, tools, and services. The company has manufacturing facilities throughout the world including in Sweden, the United States, Canada, Australia, China, India, South Africa, and Germany.

Epiroc sources steel to manufacture its equipment from a variety of locations. The steel must undergo a complex heat treatment process to achieve the right strength, hardness, and flexibility required for drill steel and other tools. A major component in steel is iron, and ingredients like metallurgical coal (carbon), manganese, cobalt, nickel, chromium, and other metals may also be part of the recipe. There are over 3,500 grades of steel, each varying in physical and chemical properties. “And different heat treatments are needed to obtain the right characteristics to ensure the highest quality," says Jonas Jern, Head of Digital Innovation at Epiroc. 

Jern says a major issue for Epiroc was that the company’s manufacturing facilities around the world were not able to easily share data and best practices for consistency in steel quality. The need for consistency differs depending on the type of tool or equipment part. Maintaining consistency in materials and operational precision across the company is paramount in everything from customer satisfaction to waste management, even safety. “This is a refined process, it’s a hard process, and takes a lot of energy.” Jern says the company was collecting massive amounts of data at its locations but had no straightforward way to fully utilize it. This created inefficiencies and redundancies, and it sometimes impacted quality—resulting in customer equipment returns. It also contributed to the organization’s inability to get precise insights into overall operations.

Epiroc wanted a central solution that could help all manufacturing locations optimize processes with the best parameters for quality, along with a way to easily assess and adjust for concerns before they occur. The solution also needed to be adaptable as the company continues to grow and innovate. Erik Nyten, Enterprise Architect at Epiroc says the company knew that AI and machine learning would play an essential role in empowering the organization to make better decisions at scale. “Making our data accessible and getting complex processes and new requirements automated in a scalable solution was key,” Nyten says.

Creating “data heaven”

Because Epiroc had already implemented Microsoft Azure across the organization, it was a natural choice to use Azure Machine Learning, Azure Data Factory, and Azure Databricks to create a solution that could deliver the precision and scalability the company needed. “And, we knew Microsoft already had policies and processes in place to help support our sustainability objectives,” says Camilla Goldbeck-Lowe, Vice President Sustainability at Epiroc. “We felt confident that with Microsoft we could create a solution to hit our key performance indicators—and help us continue to evolve and improve over time.”

The Microsoft Customer Success Team in Sweden further won the Epiroc team’s confidence by consulting with them about Azure Machine Learning and Enterprise Scale Machine Learning (ESML), an open-source solution accelerator created by Microsoft Sweden. The Customer Success Team also gave guidance on how to configure and set up an ESML AI factory to ensure concrete implementation of AI factory best practices. 

The AI factory provides a modern data architecture with Microsoft Intelligent Data Platform services, including Azure Machine Learning, Power BI, Azure Data Factory, and a data lake design to support data mesh. Using ESML, Epiroc, with the help of Microsoft Cloud Partner Program member Molnbolaget, was able to quickly establish an AI factory on Azure with secure private networking. The entire process took just 60 hours. 

"With ESML, we got a modern data architecture to support Azure Machine Learning and Automated Machine Learning for multiple use cases and projects,” says Jern. The team was also able to create machine learning models specifically for the heat treatment process with an end-to-end pipeline to the AI factory—all in just six weeks. 

With machine learning models establishing critical parameters and the ability to easily implement new AI models, accounting for specific variables and controlling process flows became easier. So did the ability to predict steel density, hardness, and flexibility for Epiroc rock drilling tools under various conditions. “We can now more accurately assess optimal tolerance levels to prevent against structure fatigue or failure,” says Nyten. The increase in efficiency and quality control in manufacturing at every Epiroc location around the world also sets the stage for greater efficiency and productivity.

Epiroc also had help from Microsoft Cloud Partner Program member Sogeti in running automation in AI governance—a crucial step in creating an enterprise scale AI factory. Nyten says it was invaluable having the assurance that Epiroc can confirm it meets the regulations and requirements of every country it operates in, and it made the roll-out process easier. “It was a team effort, but the speed and ease with which we were able to create such a large-scale solution was incredible,” adds Nyten.

Jern adds that while with the company’s old system, it was hard to access useful data or gain business insights. “Now,” he jokes, “It’s like data heaven.”

Realizing goals and applying new use cases

With the new solution, Epiroc is realizing its goal of consistent steel quality—and of reducing waste. “Now, with Azure Machine Learning and predictive AI, we consistently produce material with better quality,” says Peter Malmberg, Vice President Digitalization at Epiroc. Malmberg notes that the company has reduced customer rejections and product returns by 30 percent, saving time and money. “For us, higher quality also means fewer trips to pick up unwanted deliveries which saves the company money,” says Malmberg. “These things are wonderful, but it was really the speed of implementation that got us thinking about further innovation.”

Jerns agrees, saying the speed in which Epiroc was able to implement the solution and see noticeable improvement in the heat treatment process catalyzed his team to look at other areas of application. “Currently we have eleven analytical teams with use cases running in the Epiroc worldwide ESML AI Factory and utilizing Azure Machine Learning—with more planned in the coming year,” says Jerns. 

“Having the ability to analyze all of that data in our production process and our research and design will help us to outperform competition, accelerate innovation, and adapt to the needs and priorities of our customers,” says Malmberg. “We’ve only scratched the surface of what we can do with Azure Machine Learning."

Find out more about Epiroc on Facebook, Instagram, and LinkedIn.