The problem that I worked on was the realisation of precision temperature differences between multiple vents in the car. I worked on this project when I was a member of the french automotive supplier Valeo ( I do not share any valuable data or intellectual property). In order to achieve thermal comfort you have to have heat exchanges varying by body part. Thermal comfort in itself is very complex. Oversimplified, your feet have to be kept much warmer than your face. That means that air coming out of the vents aiming at your feet has to be hotter than air coming out of the vents aiming at your face. If you set the climate controls to 20 degrees celsius you will actually get much different temperatures at the vents. There is a standard HVAC architecture that has proven to be cheap and working.  Basically, all of the air is flowing past an evaporator to completely cool down to about 5 degrees celsius and is then separated. One part of the air then flows past a heater that is going to warm up the air to about 80 degrees celsius. After that, the warm air remixes with the cold air. The mixture can be adjusted to adjust the cabin temperature. One major problem with this architecture is that is it is really difficult to achieve the different temperatures. The conventional solution usually features special geometries in the mixing chamber that redirect and deflect hot or cold air to achieve the different temperatures.

One problem with this solution is that it is very restrictive in a sense that it dramatically increases the systems air resistence which than leads to increased noise and lower maximum heating or cooling performance. Both of which are specified tightly since noise impedes passenger comfort and maximum heating or cooling performance are important security aspects. The other problem is the very high engineering demand that the creation of such a part involves. The parts were slowly improved through an empiric trial-and-error-process where one iteration takes from one to three hours with multiple persons and expensive machinery permanently involved. It goes without saying that there is still room to improve with that process.  

The pictures above show a conventional HVAC System and the effect simple geometries can have on the system response. I started out playing a bit with features and tried to understand how exactly they were linked to system response. That actually worked out surprisingly well. In retrospect, I think a neural network should have been installed to help design and improve these features.

However, this first step was not revolutionary and more about myself diving deeper into the topic. I then tried then to come up with geometries that were not rigid, since that was an untouched degree of freedom yet. That worked somehow. The major challenge with this approach was to manage the additional kinematics needed to move the geometries, especially with the high-cost pressure in the automotive industry.

Fundamental error

While working on the topic two things really stood out to me.

First, how detached the HVAC development from the rest of the vehicle is. The car manufacturers assign space for the HVAC and leave the development entirely to suppliers. Since the entire car is a product itself which tries to perform best in its boundary conditions, the HVACs boundary conditions not only have an impact on the HVAC performance but also on the performance of all other parts and hence the entire car. But since neither manufacturers nor suppliers are capable of assigning the perfect boundary conditions, the products get worse.

Secondly, HVAC-Systems serve two general purposes. One is to ensure passenger comfort. The other is to ensure passenger safety through defogging windows or providing lower temperatures in hot climates. These two purposes impose fundamentally opposing boundary conditions on the HVAC-Systems. Trying to shoehorn them into one product is like trying to combine a leafblower with a coffeemachine so that the outcome can be sold for both usages.

I discovered these issues too late and did not respond correctly. I should have taken it a level higher and should have focused on improving HVAC development in those regards instead of trying to find a solution for the existing architecture.