PUNCH Torino accelerates hydrogen vehicle development using Oracle Cloud Infrastructure for engine design

Formerly the global center of excellence for diesel engines and electronics for General Motors (GM), the company is now a part of the PUNCH Group and has a new mission: To explore the potential to convert thermal engines from fossil fuels to hydrogen. While it continues to engineer diesel and gasoline combustion systems, PUNCH Torino sees hydrogen-powered engines as a viable alternative to electric-powered vehicles and battery technology. The company is also developing micro-mobility projects and applying AI to new fields.

“The separation from GM gave us the flexibility to apply our engineering capability to a larger, wider operation instead of what was in the past just for diesel,” says Mauro Bighi, CIO of PUNCH Torino.

Developing a hydrogen-powered engine is a complex, next-generation automotive design. To access the advanced modeling and simulation tools required for this work, PUNCH Torino uses cloud-based high performance computing (HPC) available in Oracle Cloud Infrastructure (OCI). This platform helps engineers observe and understand engine performance and emissions characteristics using computational fluid dynamics (CFD), which are compute-intensive simulations to predict flow distribution, pressure loss, heat transfer, and combustion.

With the help of OCI, the PUNCH team can design and develop engines from concept all the way through to the manufacturing process, and do so with a flexible and agile approach that supports rapid innovation. Moving its CFD simulations and virtual prototyping to OCI gave PUNCH Torino the flexibility to scale up and down, depending on the workload. Previously dedicated to a single car manufacturer at a time, the company now runs multiple segregated simulations in parallel, providing numerous automotive customers with fluid dynamics analysis without overlap.

And the stakes are high. These powertrain systems will be added to a carmaker’s vehicle design to support emission-free and zero-carbon footprint vehicles. “We are responsible for providing turnkey solutions and ensuring on-time results, leveraging our wide, cross-functional engineering skill set,” says Bighi.

Driving engine development with high performance computing

In February 2020, when the company became part of the PUNCH Group, PUNCH Torino needed to provide its 700 engineers with a new system to replace an on-premises high performance computing environment hosted by GM in Detroit. The new HPC system needed to support engineering efforts and run simulations using CFD to simulate fluid flow around various engineering models.

Following an extensive proof of concept, during which AWS, Google, Microsoft, and Oracle were invited to run multiple CFD simulations against benchmarks, PUNCH Torino recognized OCI as having the best quality, price, and performance. OCI now provides the foundation for conducting simulation and virtual prototyping services. On OCI, the workloads are running faster and require less core utilization than the on-premises environment previously used.

Partnering with HPC consulting company Do It Systems, PUNCH Torino's production environment went live just 10 weeks after completing the proof of concept. Davide Pastorino, solution architect and project manager for Do It Systems, recently talked with Oracle Senior Cloud Engineer Dimitri Manca for an episode of Oracle’s Built & Deployed series to explain the architecture.

High performance computing on OCI

Using OCI HPC, PUNCH Torino's engineers can run CPU, memory, and IO-intensive simulation and testing workloads up to 24% faster compared to the benchmark set for the proof of concept. They also use 33% fewer compute cores by managing them in an HPC cluster.

OCI provides several advantages when running CFD simulations and virtual prototyping. Engineers can scale up and down depending on the workload and run multiple simulations in parallel. OCI has also given PUNCH Torino the ability to designate the optimal Compute shape for each engine simulation, enabling engineers to balance the speed of analysis against the cost of running the compute shape over the period of the modeling cycle. According to Bighi, one of the primary benefits of moving its on-premises engine simulation system to the cloud is the flexibility to choose the compute shape.

“OCI provides incredible performance for our HPC workloads, speeding up computational fluid simulations while also optimizing costs,” says Bighi.

Architecture

In its OCI tenancy, PUNCH Torino runs the Abaqus, Converge, StarCCM+, and Optistruct applications for CFD, modeling, and simulation. The engineers access the applications through a virtual private network (VPN) from an on-premises access and control center, which is an Altair Access web application.

The control node brings up the HPC cluster nodes on demand. The control node then separates a given job into several parts and submits them to process concurrently on OCI. OCI Control Scheduler automatically scales the Compute nodes through REST APIs, and the HPC cluster deploys bare metal instances on demand. The simulations are typically optimized to complete in 5–6 hours.

PUNCH Torino uses three types of OCI storage to optimize costs: Hot, warm, and cold. Local or hot storage on the Compute nodes is used for temporary data that’s deleted at the end of each simulation. Simulation results are stored in warm file storage for ease of access during post-processing and are permanent. When an engineer has stopped working on a particular dataset, all variable data can be either downloaded or stored in the cloud within cold object storage, which can be accessed for up to eight years.

No more waiting for HPC resources

Using OCI, PUNCH Torino’s HPC workloads are all run in the cloud, and nodes are scaled based on the resources required by the simulations and the available software licenses. Overall, OCI has enabled PUNCH Torino to support high performance requirements, optimize costs, and complete critical engineering tasks more rapidly, accelerating the progress of clean engine development.

According to Bighi, using OCI delivers a major improvement over the on-premises solution of the past. “In the previous on-premises approach, if the computational resources for engine simulations were allocated, the job just had to wait until the next core was available, the next memory was available, and so on,” says Bighi. “On Oracle Cloud Infrastructure, not anymore, because the only limits that we face are the number of licenses.”

View more Oracle Connect articles