Kubota Corporation develops engines faster with high-performance computing on Azure

Innovation is rarely easy. This is especially true when engineers are on the cutting edge of design, pursuing responsible solutions for multi-fuel engines. Research, development, and testing for engines are gluttons for compute capacity, but the need for resources ebbs and flows by project and throughout the entire process.

Kubota Corporation contributes to solving social issues with food supply, clean water, and environmental conservation in more than 120 countries around the world. It manufactures engines, tractors, construction machinery, ductile iron pipes, and other products. Since starting the production of small engines for agricultural and industrial use in 1922, Kubota has produced more than 30 million industrial engines. Kubota’s engines are not only used in its own products but also as the core of various machines, working in places that people cannot usually see and supporting the world’s industries. In addition, with a lineup of 3,700 types of engines, Kubota has met emissions regulations in different countries and has maintained its position as the world’s top producer for industrial diesel engines of 100 horsepower or less.

The Engine Engineering Department at Kubota in Japan is charged with developing frontline engines for industrial machinery. Kosuke Suematsu, leader of Computer Aided Engineering (CAE) Research and Development in the department, explains, ”We assess the durability, reliability, emissions performance, fuel efficiency, and other aspects of various engines, such as diesel, gasoline, and hydrogen engines, by using numerical simulations. We predict the performance at the design stage—before making prototypes and testing them—to improve the design quality. This prevents any setbacks in development and enhances the efficiency of our development work.”

“We were developing an engine with Kubota’s new proprietary combustion system, aiming to reduce CO2 emissions,” said Suematsu. “Emissions standards have become stricter, and we needed to find design solutions that could enhance emissions performance and fuel efficiency. Traditionally, development relied on the experience and intuition of skilled engineers, but this approach often led to numerous prototypes and extensive testing, elongating the development timeline. Using numerical simulations allowed us to understand internal phenomena within the engine and swiftly generate design proposals that contributed to improving performance.” 

Simulating engine combustion requires a significant amount of computation time. Consequently, during busy periods, new calculation jobs would be queued before the previous ones finished, resulting in wait times of over a week to obtain results. Considering the need to keep engine development on track, enlarging computational capacity became critical. 

The Engine Engineering Department had traditionally used on-premises high-performance computing (HPC) systems. Upgrading the existing system was not feasible as the process of budget allocation and hardware procurement might take more than a year. These delays were unacceptable, given the Kubota development schedule. With the strain on resources and the need to speed engine development, the demand for a different technology became urgent.

Choosing a cloud-based, high-performance solution

The Engine Engineering Department considered that moving to the cloud might be an excellent solution, but there were initial concerns. Suematsu said, “One of the misgivings was the potential security risk of storing data on external servers. Fortunately, Kubota had adopted Microsoft Azure as its standard cloud platform in recent years. The company’s rules for use were well developed, and Azure was meeting the company’s essential requirements. We were able to smoothly obtain approval for its use. Azure provides virtual machines and images with the features required for HPC systems, and it also has a history of running the numerical simulation software that Kubota uses.”

Building a new environment within a limited time required advanced knowledge and skills in cloud and HPC. Altair Engineering managed the computing simulations system and provided a range of support from specification review to implementation. With the cooperation of the Kubota IT department and Microsoft, Kubota was able to build an infrastructure that was operational in less than four months. “By adopting the cloud, we were able to build what we needed in less than one-third of the time it would have taken for a different solution and meet the product development schedule,” said Suematsu.

Kubota needed a solution that gave it scale, agility, and maximum computing power, and it needed to keep a lid on costs. It found its answer by using AMD hardware, HBv3 virtual machines, and Azure. “We were already transitioning to Azure,” said Suematsu. “By using the HBv3 virtual machines that use AMD hardware, we were able to achieve high computational power and the ability to increase or decrease it as needed.”

The cloud advantage

“Since the amount of calculation involved in development varies from time to time and by season, we appreciated that this new solution could manage our changing workloads,” said Suematsu. “With on-premises servers, the initial cost increases if we tailor the computing capacity to meet peak demand. During off-peak times, however, the utilization rate decreases, leading to a diminished return on investment. The pay-as-you-go model of cloud computing aligns well with HPC needs. It allows usage as needed, with costs incurred only for actual usage. Doubling the number of compute nodes theoretically halves the computation time. While on-premises costs would approximately double, the pay-as-you-go model keeps costs equivalent. To further enhance cost-effectiveness, Kubota adopted Azure Spot Virtual Machines, which offer substantial discounts in lieu of a Service Level Agreement. In busy periods, the insufficient calculating capacity of our on-premises server could slow us down for weeks. We are grateful for the pay-for-what-you-use cloud model.“

The original proof of concept showed positive results in computation time, business agility, cost optimization, and scale of development. Compared to the on-premises system, the company scaled up its parallel computing by more than three times at less cost and sliced combustion simulation time from 15 to 7 hours. Where once engineers had to wait for over a week for their calculations, now they can submit their jobs before leaving work and check the outcomes the next morning.

“When computational needs vary over time, we believe our cloud-based system is less than half what the cost for a new on-premises server would be,” said Suematsu. “Maintenance, updates, and other secondary work has largely disappeared from our team, meaning we can keep our focus on our main jobs.”

Kubota is currently using numerical simulation software tools such as Siemens Simcenter STAR-CCM+ to develop engines that accommodate decarbonized fuels, including hydrogen. It has high expectations that these new computational resources will pave the way for creating more efficient and less costly engine designs in the future.

“By rapidly establishing a flexible and powerful HPC environment on Azure, we’ve unlocked the ability to explore ideas that were previously constrained by computation-time limitations,” said Suematsu. “Moving forward, we remain committed to addressing the diverse needs of our customers by developing industrial engines and contributing to the resolution of new societal challenges, such as achieving carbon neutrality.”

Find out more about Kubota Corporation on Facebook and LinkedIn.