Abstract
Aiming at the problem that it is difficult to model and analyze the thermal effect of large motors as a whole, a method based on fluid network decoupling is proposed to solve the global thermal effect of large motors. Taking YJKK500-4 2500-kW compact medium-sized high-voltage motor as an example, combined with the internal structure and the actual size of the motor, and using the flow-heat coordination mechanism and the fluid network, the whole internal ventilated paths of motor are divided into four equal pressure regions: end, stator-rotor, end, and internal fan according to the magnitude of pressure. The reasonable basic assumptions and boundary conditions are selected to link each region, which can avoid the problem that the inlet boundary condition is difficult to give. After that, the internally ventilated paths are modeled and analyzed, and the fluid flow and temperature in each region of the internally ventilated paths are obtained. Combined with the experimental data, the calculation results are all within the error. The rationality and correctness of the method used in the paper are confirmed, which solves the problem that the ordinary computer cannot calculate the huge model of large motor and provides a basis for the calculation of similar motor in the future.
Introduction
Designing efficient cooling systems for large, high-voltage motors, such as the YJKK series, presents a significant computational challenge. These motors utilize intricate internal ventilation paths where fluid flow and heat transfer are deeply coupled. Traditionally, researchers have struggled with a "Goldilocks" dilemma: simulating the entire motor as a single model is often computationally prohibitive for standard hardware, while simulating isolated sections frequently fails due to the difficulty of defining accurate internal boundary conditions, such as air pressure and velocity. This research addresses this gap by proposing a method that is both computationally feasible and physically accurate.
Methodology: Fluid Network Decoupling
The core of this study is the Fluid Network Decoupling approach. Rather than attempting a monolithic simulation, the motor’s internal ventilated paths are divided into four "equal pressure regions" based on internal pressure magnitudes: the inlet-side end region, the stator-rotor region, the outlet-side end region, and the internal fan region.
This decoupling strategy allows the output of one region to serve as the boundary condition for the next, effectively solving the "unknown boundary" problem. The study employs the Navier-Stokes equations for momentum and the RNG $k-\epsilon$ turbulence model to handle high Reynolds numbers. Numerical convergence is accelerated using the SIMPLEC algorithm. Furthermore, on-site measurements of motor losses (copper and iron losses) are integrated as heat source inputs, creating a robust field-circuit combination model.
Results and Discussion
The proposed method was validated using a YJKK500-4 2500-kW prototype. The simulation revealed critical insights into the motor's thermal behavior. For instance, the rotation of the rotor ventilation channel acts as an auxiliary fan, significantly accelerating fluid flow on the rotor side.
A notable observation was the temperature asymmetry between the upper and lower windings. Fluid entering the stator ventilation ditch from the air gap moves faster, providing superior cooling to the initial surfaces it contacts. The model also identified fluid vortexes caused by sudden structural deformations in the ventilation paths, which decrease cooling efficiency and increase noise.
Quantitatively, the theoretical maximum temperature for the stator winding was calculated at 115°C, compared to an experimental measurement of 120.1°C. This 4.4% error margin confirms the high accuracy and practical utility of the decoupling method for engineering applications.
Conclusion
The fluid network decoupling method successfully enables high-fidelity thermal modeling of large motors on standard workstations. While theoretical results were slightly lower than experimental data—likely due to modeling simplifications such as ignoring axial initial velocity in the air gap—the approach provides a solid foundation for future design optimizations. Future work will focus on refining boundary assumptions and smoothing structural transitions to minimize vortex formation and further enhance motor power density.
Summary updated