Experimentally Validated Comparison of Two CFD Simulation Workflows for Oil Jet Cooling and LPTN Improvement

Modern electric vehicles rely on powerful motors that must be incredibly compact and efficient to drive the wheels. However, as these machines pack more power into smaller spaces, they generate intense heat. If this heat is not removed quickly, the motor can fail or lose performance. For the first generations of electric cars, engineers used water jackets that cooled the outside of the motor, but this method is like trying to cool a hot cup of coffee by blowing on the rim; the heat from the center takes too long to escape. To solve this, the industry is moving toward direct cooling, where oil is sprayed directly onto the internal components, specifically the copper wires at the ends of the motor. This oil acts as a non-conductive fluid that carries heat away rapidly, allowing the motor to run hotter and faster without breaking. The challenge for engineers is predicting exactly how well this oil will cool the complex, tangled shapes of the wires, a task that requires sophisticated computer simulations to avoid building expensive prototypes that do not work.

A team of researchers from the University of Navarra and Dauch Corporation set out to test two different ways of simulating this oil cooling process. They wanted to know which computer method could predict the temperature of the motor most accurately while using the least amount of computing power. To find the answer, they first built a physical test model. They took a real motor stator, which is the stationary part of the motor containing the copper windings, and placed it inside a clear plastic container. This container allowed them to watch the oil flow while keeping it contained. They ran electricity through the wires to generate heat, just like a real motor would, and used precise sensors to measure the temperature at various points. They tested the system with oil flowing at two different speeds and at two different starting temperatures, creating a reliable set of real-world data to compare against their computer models.

The researchers then ran two distinct types of computer simulations to see how well they matched the physical experiment. The first method, known as the Finite Volume Method, breaks the space around the motor into millions of tiny, fixed boxes. It tracks how the oil and air move through these boxes, calculating how the liquid drips and spreads over the wires. This approach is very detailed and versatile, capable of showing exactly how the oil interacts with the solid metal, but it is extremely slow. The second method, called Moving Particle Simulation, does not use fixed boxes. Instead, it treats the oil as a collection of individual particles moving through space. This approach is much faster because it does not need to calculate the empty air around the motor in the same way, but it requires a separate step to figure out how the heat moves inside the solid metal wires.

When the team compared the results of both simulations to their physical measurements, they found that both methods were remarkably accurate in predicting the average temperature of the oil-cooled wires. The detailed, box-based method matched the real-world temperatures almost perfectly, while the particle-based method was also very close, though it occasionally struggled to predict the very hottest spots when the oil flow was slower. The particle method proved to be significantly faster, taking only a few hours to run on a powerful graphics card, whereas the box-based method took days to run on a large server. This speed difference is crucial for engineers who need to test many different designs quickly. The researchers also discovered that the particle method was actually better at predicting the temperature on the side of the motor that was not being sprayed with oil, likely because the other method simplified the metal structure too much to be fully accurate.

The most significant finding of the study was what happened when the researchers took the detailed heat data from these simulations and fed it into a simpler, faster model used for everyday design work. This simpler model, known as a thermal network, usually relies on rough guesses about how well oil cools the motor. When the team replaced those guesses with the precise data from their advanced simulations, the simple model became incredibly accurate. It transformed from a tool that could be off by dozens of degrees into a reliable predictor that matched the real-world tests. This means that engineers can now use fast, simple models to design cooling systems with the confidence of a complex, slow simulation, provided they first calibrate them with high-quality data. The study concludes that while the slower, detailed method offers great flexibility, the faster particle method is an excellent choice for simulating oil jet cooling, offering a powerful new way to design the next generation of electric motors.