Thermal modelling of electrical machines

For the thermal modelling of electrical machines, three main methods exist and are used by EOMYS [1]: the lumped method, the finite element modelling and CFD simulation. EOMYS has an expertise in the three methods as each method is complementary with the others. They all have different levels of accuracy and different computational times. Moreover, each method can be used in 2D or 3D.

Lumped Method

The lumped method (also called Nodal Method or Thermal Network Method) is based on electrical resistances analogy to simulate the thermal behavior of the machine. The model is macroscopic: results will be obtained for elements of the machines (stator teeth, windings, yoke, etc.) instead of having local results. This method can be used in steady or unsteady states with variable external loads (for example, a driving cycle for a car motor). The main advantage of the method is its low computational time: it’s the ideal method for optimization. However, it requires the convective boundary conditions inside the machine. They can be determined using correlations from scientific literature, experimental results or more complex CFD simulations.

<span lang='en'>Modelling of a 6 poles machine with Lumped Method</span>
Modelling of a 6 poles machine with Lumped Method

Conductive simulation using Finite Elements Method

Finite Elements Method (FEM) can solve directly the heat equation inside the solid parts of an electric machine. With this type of modelling, the local temperature field can be evaluated: it allows a detailed analysis of hot spots. The mesh of an electromagnetic simulation can be directly used for the thermal modelling and the two resolutions can easily be coupled. The computational time is higher than the lumped method. However, the convective boundary conditions must be imposed like the lumped method: correlations or CFD results must be used.

<span lang='en'>2D FEM Simulation: cooling of a stator with a water channel</span>
2D FEM Simulation: cooling of a stator with a water channel

CFD Modelling

Computational Fluid Dynamics (CFD) allows a resolution of the flow inside the machine coupled with heat transfer. The required mesh are generally very fine and may depend on the fluid velocity. Computational time is very high. With CFD, there is no need to use correlations from literature the coupled velocity and thermal fields are directly solved. The level of uncertainties is, however, relatively important and the convergence of the calculations is not necessarily guaranteed especially for very unstable flows (high turbulence level, high rotating speed, etc.). A good practice for CFD in electrical machines is to focus on certain elements of the machine and use the results (flow rates, heat transfer coefficients, etc.) in a FEM or lumped model of the full machine. For the cooling of electric machines, CFD is usually done in steady state: the time to reach the steady state of the flow is much lower than the thermal characteristic time, which can reach a few hours.

<span lang='en'>Flow and heat transfer coefficient in a rotating channel</span>
Flow and heat transfer coefficient in a rotating channel
<span lang='en'>Flow distribution and temperature field in a stator with cooling channels</span>
Flow distribution and temperature field in a stator with cooling channels

Lire aussi

  • Scientific software development
    EOMYS R&D engineers have strong skills in software development and IT project management. Combined with their skills in applied physics, applied mathematics and scientific computing, this enables them to implement and validate robust, easy to  (...)
  • MANATEE e-NVH software
    To support its consulting activities, EOMYS has progressively developed a fast multiphysic (electromagnetic & vibro-acoustic) simulation software of electrical machines, called MANATEE - Magnetic Acoustic Noise Analysis Tool for Electrical  (...)
  • Optimization of complex systems
    EOMYS implements the latest applied mathematics tools in order to find the global optimal design of a system, or post-process a large amount of data (e.g. experimental data of a prototype) to extract the relevant information.
    Every design  (...)