DIY PWM PID DSC Cookie Oven

For 24-451: Feedback Control Systems

Overview

One classic application of DSC (dynamic systems control) is the oven. In order to study and optimize this system, we downsized to build an oven that fit only one cookie, manufactured it, built the wiring harness, and optimised a PID (proportinal, integral, derivative) control system to hit various control criteria.

Physical Construction

Our oven is machined out of a scrap piece of aluminum

The material was bored out on a lathe, the walls were thinned out, and the thermocouple and heating pin holes were added on the mill

Final assembly includes the two heating pins, the two thermocouples, the two oven halves, the baking tray, and the heat sheild for the top pin.

Electrical Setup

  • We used two k-type thermocouples to sense the temperature in the oven. Thermocouples are cheap and easy to read, and k-type thermocouples in particular matched our intended temperature range.

  • We used 2x 500W cartridge heaters to heat the oven. This proved to be way more than was necessary. They plugged directly into the wall, but the power was toggled by two Arduino controlled relays.

  • The relays allowed us to control wall voltage heaters with arduinos, which normally only power up to 5 volts.

  • An Arduino Uno runs all of the code, which mainly contains the control scheme to take in thermocouple voltage, convert to temperature, and create an output for the relays

  • This is our initial control scheme, which simply uses bang-bang control to pass full power to the heaters when the oven is colder than intended, and turn the heaters off otherwise. Bang-bang control is good for simplicity, but it does not perform well with other control metrics.

  • This is the system response to the simple bang-bang control scheme. It overshoots dramatically, and is very underdamped. Overshoot 28% | Peak time 250s | Settling time 680s | Steady state error 1% - 7%

  • This is the proportional control scheme. For most of the range of error, it also returns full power or no power to the heaters. However, in the control range the affective voltage passed to the heaters via PWM. The colder the oven gets, the more the duty cycle is biased towards full power. This helps to prevent oscillations and overshoot.

  • This is the system response given the proportional control scheme. It does have more damping and less overshoot, but it also gains a significant amount of steady state error. Overshoot 0% | Peak time 100s | Settling time 300s | Steady state error ~12%

Takeaways

  • Real world control schemes are difficult. You can do lots of analysis to work out possible solutions, but the simulated responses are only as good as your system models.

  • Multiple time constants can be difficult to work with. The oven could be actively heated, but only cooled naturally. This meant that overshooting the set temperature was very costly and we had to lean on the side of being below temperature or wait a long time to regain control

  • Physical components play a key role. We spent a lot of time trying to optimize the control scheme given our physical set up, but ultimately one of the most powerful tools we had was to adjust the physical set up itself. By thinning out the oven, and using relays with faster switching speeds, we were able to improve time constants and tighten up the response drastically.

Future work

As can be seen on the right, the oven, using the proportional control scheme, works! It is actually able to go from cold to cooked cookie in about 12 minutes. However, there are still improvements that could be made. Namely, the oven does not accurately rest at the set temp, and could get to the set temp faster. These could be accomplished with the addition of integral and derivative terms. The integral term will help with the steady state error issue, but could add to more overshoot. That is where the derivative control comes in. The final step would be to tune the three relative to each other for maximally efficient cookie production.