Much faster and more efficient: Digital simulations at Schaeffler with Azure HPC

At Schaeffler, experiments with prototypes are a thing of the past. The motion technology company now relies on digital simulations to examine how a component or system will behave under certain conditions. Recently, there has been such huge demand for these simulations that the old on-premises landscape simply no longer had sufficient computing capacity. Help arrived in the form of high-performance computing (HPC) in the cloud. Using a new solution based on Azure HPC and some software provided its partner Rescale, Schaeffler is running more simulations than ever before—and with unprecedented speed and efficiency.

The challenge: High manual effort and long computing times slow down simulations

Whether it’s wheel bearings for cars or rotor shafts and gears for wind turbines—each Schaeffler product begins life in the company’s research and development departments. Twenty years ago, lab experiments and prototype arrays were the order of the day. But as the company’s digital transformation progressed, these were largely replaced by digital simulations. That’s how developers analyze, calculate, and test the likes of load behavior and heat generation for an individual ball bearing, or how these aspects will affect an overall system. “Simulations allow us to perform multiple experiments simultaneously, which leads to faster results and shorter development times,” says Markus Kießling, Product Owner – Systems Engineering Solutions at Schaeffler.

Models that require a lot of computing power can be worked on in a variety of applications, such as MATLAB, Ansys, and Abaqus. “These applications contain more than 120 tools for the various types of simulations, with new versions being released every year. Since our projects have to use the same version from beginning to end, we’ve always got multiple versions running at the same time,” Kießling says. “And if these tools are installed locally on our computers, we have to spend a lot of time updating, licensing, and managing each one.”

That’s why Schaeffler began by providing applications through an on-premises data center. “We bought and installed our own hardware. The application operators then installed the tools,” says Tobias Frömel, Product Owner – Hybrid Cloud Platforms at Schaeffler. “This established an extremely isolated infrastructure that took considerable manual effort to maintain, and making it cost-effective meant having it in place for a longer period of three to five years.”

When the number of digital simulations skyrocketed and the on-premises solution was reaching its limits, the various departments began setting up and running their own solutions. “Our aim was to bundle everything within Corporate IT and give the whole thing a new footing. The only way forward was to use an HPC cloud solution because this ensures we always have the latest and most efficient HPC infrastructure at our disposal,” Frömel says. “Our developers need to be able to run a host of simulations securely, quickly, flexibly, and at the same time. Azure HPC creates the perfect basis for that.” As Schaeffler had already been pursuing a cloud-first strategy for several years with Microsoft as its strategic partner, it chose this product as its technological basis and the team began the rollout.

The solution: Azure HPC raises simulations to a whole new level

Today, the new standardized solution brings together 20 to 30 of the previous individual solutions and is made up of two parts. One is the front-end web portal that runs via a Software as a Service (SaaS) solution from the partner Rescale and resembles an app store where developers can select the appropriate simulation application. The other runs in the background and is made up of the various Azure compute resources and Azure Storage services for interim data storage.

The solution is fully automated: as soon as a simulation is started in the front end, the precise amount of storage resources and computing power required are immediately made available in the back end and vacated once the simulation is complete. Users can choose to run their simulations either in the web portal or at a virtual workstation in the cloud. “Both options no longer require any additional installation. It takes just a few clicks to configure the system for a given simulation,” Frömel explains. “This is hugely advantageous in terms of administration, user-friendliness, and operability—something our colleagues really appreciate. Through this project, we’ve gained a better understanding of their needs and can now work with them to develop more custom solutions.”

Thanks to the cloud and automation, running a simulation uses only the capacity and resources that are required. “In other words, we pay only for what we actually use while getting the maximum in flexibility and scalability the cloud has to offer,” Frömel says. This means that Schaeffler can now run 12,000 simulations for a large-scale project in a matter of days. Since the calculations are performed in the background, users can continue to access all applications as usual. “Thanks to our new HPC solution based on Azure and Rescale’s SaaS solution, we can complete our simulations a good deal faster,” Kießling says. “This means we can begin the next round of calculations sooner, detect errors earlier, and develop products far more quickly. That’s a massive boost to our competitiveness and innovative strength.” Meanwhile, the shorter computing cycles have a direct effect on costs because application licenses are often invoiced according to duration and frequency of use. What’s more, as Schaeffler’s research and development activities involve proprietary data that must be protected, the company benefits from Microsoft’s security framework.

Due to the high level of compatibility of Azure, Schaeffler will have no problems later on when expanding the solution to include partner systems or solutions developed in-house—regardless of whether they run on-premises or in the cloud. “Thanks to the HPC solution based on Azure, we now have the perfect foundation from which to make it considerably easier for our developers to access future innovations like quantum computing,” Frömel says.