A Hybrid Model for Fast and Efficient Simulation of Fluid Power Circuits With Small Volumes Utilizing a Recurrent Neural Network

• Published in: IEEE access 2022, Vol.10, p.48824-48835
• Main Authors: Ustinov, Stanislav, Wu, Huapeng, Handroos, Heikki
• Format: Article
• Language: English
• Subjects: Computational modeling; Differential equations; Dynamic simulation; Fluid power; fluid power system; Fluids; Integrated circuit modeling; Mathematical analysis; Mathematical models; Model accuracy; Neural networks; Numerical models; Power systems; Real time; real-time simulation; Real-time systems; recurrent neural network; Recurrent neural networks; Simulation; small volumes; Systems simulation; Virtual prototyping
• ISSN: 2169-3536; 2169-3536
• DOI: 10.1109/ACCESS.2022.3172662

The modeling and simulation of fluid power systems are essential parts of the real-time simulation of virtual prototypes of mobile working machines. In several cases in the dynamic simulation of such fluid power systems, a longer simulation time is required. This makes the traditional mathematical models inefficient for real-time simulations, particularly when simulating fluid power systems because of the small volumes in stiff differential equations of pressure. To overcome this issue, a novel hybrid model is proposed for stiff fluid power systems simulation. The main feature of the model is the utilization of a recurrent neural network instead of stiff differential equations of pressure with small volume. At the same time, the dynamics of the rest system are traditionally presented with algebraic and differential equations. The testing results of the introduced hybrid model showed that the novel method can reduce the simulation time, which makes the model suitable for real-time applications. Moreover, the accuracy of the model remains at a fairly high level compared to traditional mathematical models.