A major recognition for the AIT Austrian Institute of Technology and its partners: The journal ISA Transactions has awarded the paper "Hybrid control of hydraulic directional valves: Integrating physics-based and data-driven models for enhanced accuracy and efficiency" the ISA Transactions Best Paper Award 2025. Presented annually by the Editor-in-Chief and Elsevier, the award recognizes the journal's most outstanding scientific publication of the year.
The award-winning research was carried out in collaboration between the AIT Austrian Institute of Technology, Bosch and TU Wien. It demonstrates how physics-based models and data-driven methods can be combined into a powerful hybrid control strategy for hydraulic directional valves.
When Physics and AI Work Together
Hydraulic directional valves are key components of modern construction, agricultural and mobile machinery. They control the flow of hydraulic fluid and determine how components such as excavator arms, crane booms or loader buckets move—in which direction, at what speed and to which position. To improve energy efficiency, many of these valves are designed to seal almost completely when in their neutral position. While this minimizes energy losses, it also makes precise control more challenging: Before hydraulic fluid can begin to flow, a so-called dead zone must first be overcome—a range in which the valve does not yet respond despite being actuated. In addition, only limited measurement data, such as the valve position, are typically available in industrial applications, while important variables such as the internal pressure conditions cannot be measured directly. The award-winning paper demonstrates how these challenges can be addressed by intelligently combining physics-based models with data-driven methods.
Machine learning does not replace physics—it complements it.
The researchers developed a hybrid control strategy that combines the strengths of both approaches. A physics-based model describes the valve's behavior according to physical principles, while a data-driven model learns those complex effects that are difficult to capture mathematically. This keeps the control system robust and highly accurate while significantly reducing the effort required to model complex valves in full mathematical detail.
Experimental validation on an industrial test bench demonstrates that the hybrid controller achieves the same high control accuracy as the conventional physics-based approach. At the same time, the control strategy can be transferred more easily to different valve variants, reducing development effort and enabling manufacturers to introduce new valve designs more quickly—an important advantage for the continued development of modern hydraulic systems.
Research Delivering Direct Benefits for Industry
With the Best Paper Award, ISA Transactions recognizes scientific publications that make outstanding contributions to automation and control engineering. The award highlights the research approach pursued by AIT and its partners: combining machine learning methods with physics-based models to develop robust, practical solutions for industrial applications.
Publication:
Tobias Glück, Amadeus Lobe, Adrian Trachte, Matthias Bitzer, Wolfgang Kemmetmüller. Hybrid control of hydraulic directional valves: Integrating physics-based and data-driven models for enhanced accuracy and efficiency. ISA Transactions, Volume 157, 2025, Pages 280-292, ISSN 0019-0578,
doi.org/10.1016/j.isatra.2024.12.029
www.sciencedirect.com/science/article/pii/S0019057824006153
ISA Transactions:
www.sciencedirect.com/journal/isa-transactions/about/news/congratulations-to-the-winners-of-the-isa-transactions-best-paper-award-2025