Designing a Formula SAE Electric Cooling System
This year I was tasked with designing the cooling system for F28, one of my FSAE team's first electric racecars. This is by far one of my largest projects; it took ~3 months not including manufacturing and testing. I researched designs other teams use, wrote a cooling simulation to determine radiator sizing, modeled the cooling system, worked with my teammates to integrate it into the rest of the car, and presented my designs to the team in preliminary and critical design reviews. I learned a lot about heat transfer, fluid mechanics, MATLAB, Creo, GD&T, and working in a team environment when doing this project.
 |
| F27, the team's last combustion car |
At the time of writing, I manufactured and assembled parts of the system but the COVID 19 pandemic just started so we will likely be finishing the car and test the cooling system this coming fall.
 |
| F28's Cooling System |
I started the design of the cooling system by doing research on what other teams did and learning the theory behind what I was designing. Since this is one of our first electric cars, I didn't have any past cars to base the design off of. I read a few chapters of a heat transfer textbook that I borrowed from one of the older members on the team. When planning I allocated a lot of time to the research; I haven't taken heat transfer yet so most of the theory was completely new to me. I also made a list of relevant FSAE rules for the cooling system in order to stay rules compliant for competition.
 |
| Cooling loop diagram |
I determined that a single loop cooling system was the best option for cooling both the motor and the motor controller. The advantage of a single loop cooling system is simplicity; in addition to being lighter, it's much easier to design and build one loop rather than a separate system for the motor and motor controller.
A major part of this project was building the cooling simulation to determine the size of the heat exchanger(s) / radiator(s). By using the NTU method of heat transfer I was able to get a reasonable idea for the heat lost to the air based on a function of flow rate, airspeed through the radiator, ambient temperature, water inlet temperature, and radiator size. I was lucky to have a heat transfer coefficient for the radiator from past team data.
Calculating the heat into the system was surprisingly simple, with a generous assumption that all the inefficiencies in the motor and motor controller are cooling load it's easy to get a value for the heat added into the system based on the current torque and rpm of the motor. I also used a factor of safety of 1.5 to account for the many unpredictable variables.
Using MATLAB to convert the efficiency map from the motor datasheet into a 5th-degree polynomial approximation
With a function for the heat added and rejected from the system, it wasn't super difficult to write a cooling simulation. I used data from past cars and predicted lap time sim data as an input to the cooling simulation. I ended up selecting one medium-sized radiator based on the results of the simulation. The simulation code is available here.  |
| Temp needs to stay below 50C. The dip in the middle is the driver switch in the middle of the 22km race. |
 |
| Total pressure drop for speccing pumps |
 |
| Tubing and fitting pressure drops |
With a good estimate for the pressure drop at 8L/min, I was able to spec out sufficiently powerful and lightweight pumps. I also specced out a radiator fan, tubing, fittings, and temp sensors. I included the temperature sensors in the design so that when designing the cooling system next year, we have temperature data that we can use to optimize the cooling simulation get a lighter radiator.
The next step in the design process was modelling everything in PTC Creo and integrating it into the rest of the car. I learned a lot of Creo modeling and also learned how to use Windchill, the Creo PDM.
For the most part, there weren't many clearance issues other than the pumps intersecting with the batteries. Luckily I was able to catch this interference pretty early and work out a solution with my teammate before it became a problem. I also worked with the aerodynamics lead on designing the radiator mounting to make sure it fit well mechanically but still satisfied the aerodynamic drag and mass flow rate requirements.
 |
| Cooling system with the motor and motor controller |
 |
| Resolving clearances early in the design is extremely important |
I made an FMEA and a full BOM for my system for the FSAE rules and the competition cost event. Though this was a bit tedious, I think it's a good representation of what I would need to in a real-world engineering project.
I also made drawings for all the machined parts in the system. As a team, we manufacture all of our parts in our own shop and use drawings to document designs for when people graduate, so it was important to follow all the proper GD&T guidelines for my parts. This was pretty easy for the cooling system since there were no complicated machined parts involved.
 |
| An example of one of the drawings I made for the cooling system |
The last part of the design was making a preliminary and critical design review and presenting the design to the team. After a lot of late nights I was able to get the PDR done as scheduled, and fix all the necessary changes before presenting the CDR 2 weeks later. Both design reviews are available here.
Update: the car is now driving, functional, and not overheating! Here is a picture of the back half of our car right after I finished installing and filling the cooling system:
 |
| Just installed |
 |
| Slightly revised design with a larger swirlpot |
The only part of the design that changed was a larger swirlpot, the turbulent water, and low water level in the short swirlpot resulted in bubbles being introduced into the system. Bubbles are unpredictable and can greatly reduce heat transfer and cooling capabilities, they also risk damaging the pumps. In my initial calculations, I only accounted for the thermal expansion of the water, not the vortices forming as it's sucked out of the swirlpot.
Preliminary data largely agrees with the results of my simulation which bodes well for choosing an even smaller radiator for next year's car.