A GPU-boosted high-performance multi-working condition joint analysis framework for predicting dynamics of textured axial piston pump

Xin Yao, Yang Liu, Jin Jiang, Yesen Chen, Zhilong Chen, Hongkang Dong, Xiaofeng Wei, Teng Zhang, Dongyun Wang

Abstract

Accurate simulation of axial piston pump (APP) dynamics is essential for design and maintenance. However, traditional CPU-based solvers struggle with the refined meshes required for textured surfaces. This paper presents the GPU-boosted high-performance Multi-working condition joint Analysis Framework (GMAF), which utilizes a Preconditioned Conjugate Gradient (PCG) method with an Approximate Symmetric Successive Over-Relaxation (ASSOR) preconditioner. GMAF leverages massive parallel computation by solving joint algebraic systems, achieving significant speedups and revealing how micro-textures improve pressure capacity and torsion resistance.

Introduction: The Challenge of Simulating Textured Surfaces

Axial piston pumps are critical components in heavy machinery, transmitting power under extreme conditions. Improving their efficiency often involves adding micro-scale "textures" to the piston surface to enhance lubrication. Simulating these features requires solving the complex Reynold’s equation on extremely refined meshes, often exceeding millions of degrees of freedom.

Traditional sequential CPU solvers are prohibitively slow for such tasks. While GPU acceleration offers a solution, standard implementations often suffer from low computational intensity when handling many small, sequential simulation tasks. The GMAF framework was designed specifically to overcome these bottlenecks by maximizing GPU utilization.

Methodology: The GMAF Framework

The core of GMAF is the Joint Algebraic System. Instead of solving for individual working conditions sequentially, the framework bundles multiple conditions (typically nine) into a single large system. This approach increases the computational load per kernel launch, ensuring the GPU's thousands of cores remain active.

To solve this system efficiently, GMAF employs a Preconditioned Conjugate Gradient (PCG) solver with the Approximate Symmetric Successive Over-Relaxation (ASSOR) preconditioner. Unlike traditional SSOR, which has recursive dependencies that hinder parallelization, ASSOR is designed for high compatibility with GPU architectures.

Convergence performance of different solvers and optimization of the relaxation factor.

As shown in the figure above, the ASSOR II variant demonstrates significantly faster convergence (lower residual over fewer iterations) compared to the standard Jacobian or SSOR methods. Furthermore, the framework's performance is optimized by selecting an ideal relaxation factor, typically around 1.5, which minimizes the total number of iterations required for convergence.

Key Results and Performance Benchmarks

The GMAF framework was benchmarked against traditional CPU solvers and standard GPU acceleration strategies. The results indicate a massive leap in efficiency, with GMAF providing speedups of 124.72% to 520.05% over sequential GPU methods.

Comparison of computational time and acceleration ratio between CPU and GPU solvers across varying mesh sizes.

The scaling behavior is particularly noteworthy. As the number of discretized points increases toward 3 million, the CPU's computational time grows rapidly. In contrast, the GPU's time remains relatively stable after an initial overhead jump. This leads to an acceleration ratio that exceeds 8x for large-scale problems, proving that GMAF is uniquely suited for the high-fidelity meshes required for textured surface analysis.

Physical Insights: The Impact of Surface Texturing

Beyond raw speed, GMAF enables detailed physical analysis. The simulations revealed that textured surfaces create distinct "steps" in the pressure field, which directly contribute to improved load-carrying capacity and torsion resistance.

The study also identified a decoupling of forces:

  • Normal Pressure Forces: Reach a steady state almost immediately.
  • Viscous Shear Stress Forces: Evolve gradually over multiple cycles.

This distinction is vital for engineers predicting long-term wear and fatigue in hydraulic systems.

Conclusion and Future Directions

GMAF represents a significant advancement in hydraulic machinery simulation, successfully bridging the gap between high-fidelity modeling and practical computation times. While the current synchronized convergence strategy is highly effective, future work will explore optimizing CPU-GPU communication to enable even faster asynchronized solving. Additionally, the researchers plan to expand the framework to include full thermo-elastic-hydrodynamic (TEHD) coupling to account for thermal deformation in high-heat environments.

Summary updated

Abstract

Accurate simulation to dynamics of axial piston pump (APP) is essential for its design, manufacture and maintenance. However, limited by computation capacity of CPU device and traditional solvers, conventional iteration methods are inefficient in complicated case with textured surface requiring refined mesh, and could not handle simulation during multiple periods. To accelerate Picard iteration for predicting dynamics of APP, a GPU-boosted high-performance Multi-working condition joint Analysis Framework (GMAF) is designed, which adopts Preconditioned Conjugate Gradient method (PCG) using Approximate Symmetric Successive Over-Relaxation preconditioner (ASSOR). GMAF abundantly utilizes GPU device via elevating computational intensity and expanding scale of massive parallel computation. Therefore, it possesses novel performance in analyzing dynamics of both smooth and textured APPs during multiple periods, as the establishment and solution to joint algebraic system for pressure field are accelerated magnificently, as well as numerical integral for force and moment due to oil flow. Compared with asynchronized convergence strategy pursuing local convergence, synchronized convergence strategy targeting global convergence is adopted in PCG solver for the joint system. Revealed by corresponding results, oil force in axial direction and moment in circumferential directly respond to input pressure, while other components evolve in sinusoidal patterns. Specifically, force and moment due to normal pressure instantly reach their steady state initially, while ones due to viscous shear stress evolve during periods. After simulating dynamics of APP and pressure distribution via GMAF, the promotion of pressure capacity and torsion resistance due to textured surface is revealed numerically, as several 'steps' exist in the pressure field corresponding to textures.
BibTeX
@misc{yao2025gpu,
  author = {Xin Yao and Yang Liu and Jin Jiang and Yesen Chen and Zhilong Chen and Hongkang Dong and Xiaofeng Wei and Teng Zhang and Dongyun Wang},
  title = {{A GPU-boosted high-performance multi-working condition joint analysis framework for predicting dynamics of textured axial piston pump}},
  year = {2025},
  eprint = {2511.06824},
  archivePrefix = {arXiv},
  url = {https://arxiv.org/abs/2511.06824},
}
Figure