Software System
Software
GitHub
We set out to have an easy method of system control. Ideally, anyone could walk up to our recycler and, through interacting with our GUI, understand how to use it. That of course meant that we had to build a GUI which would be intuitive to use and which would connect to our electrical system. To do this, we used the Python package PyQt, which offered a more updated, and therefore more professional, look than other packages such as Tkinter. It also was easier to implement than some other packages, which meant that the integration process would be less painful. The result was a GUI which walks the user through the entire recycling process, and which records data about the recycling job so that anyone can view the job log to assess system performance over time.
Firmware
The firmware of ReFilament is defined as all the embedded code written for Arduino that the user doesn’t interact with directly, but manage the spooling control and heating control systems and communicate that data with the Raspberry Pi. It can all be found here.
The control code for the spooling system required a pretty straightforward output, just analogWriting a PWM pin that controls the FET driving the spooling motor. However, the sensor input to the system is actually quite complicated. It involved reading a 24 bit dataframe from the calipers asynchronously. The way we do this is in the decode() method in ReFilament.ino. When the clock line from the caliper level shifter triggers an interrupt on the falling edge, it sets clockFlag to 1, which signals to the main loop to read the current value on the data pin from the caliper level shifter. The reason the interrupt only sets a flag is to avoid an interrupt upsetting the control algorithm or the Serial communication flow, instead waiting until the next availability in the main loop to process the caliper data.
Once the caliper data has been read, it also has to check whether the entirety of the caliper dataframe has been read or not, or if the value should be negative. This happens in all of the lines after “now = millis()” in decode().
At this point, the caliper measurement is fed into the spooling motor control feedback loop. If the filament is too thin, the spooling motors are slowed down. If it is too thick, the spooling motors are sped up. In its current iteration, the spooling motors are driven at a constant speed due to a lack of filament cooling consistency, preventing us from tuning the control algorithm.
The heating control code for ReFilament’s barrel heater coil also has a very simple input and a complicated output. The input takes the form of a negative temperature coefficient thermistor fed through a Wheatstone bridge, which in the end is just an analogRead(0) in the Arduino code. The output however, takes some processing. During each iteration of the main loop, the analog output of the thermistor circuitry is read, and the temperature is calculated by a linear approximation of the thermistor transfer function, chosen to be accurate in the range of 180-220 degrees Celsius. The loop then calls update_temp(), which is a simple thermostat function, turning the heater coil on if the temperature is below 200 degrees and turning the heater coil off if the temperature is above 210 degrees.
Serial communication was chosen as the data transfer protocol of choice for our project. between the Raspberry Pi and the Arduino. At a high level, at the press of the GUI preheat button, the Raspberry Pi tells the Arduino to turn on. After preheat finishes, the Raspberry Pi tells the Arduino to turn the motors on. On the other hand, the Arduino sends back a packet containing the current temperature of the nozzle, and the current filament diameter read by the calipers.
In terms of serial setup, we choose to run at a baud rate of 9600 bits per second. Although many people opt to use use baud rates greater than 50000 bits per second, 9600 bits per second is more than enough for the data transfer bandwidth of our project. On the Raspberry Pi, we opt to use PySerial (a python serial library). To make our data communication reliable, we made a function called “getSerialData()” which reads data from the serial buffer until termination at a new line. The use of a function also means it’s expected behavior always stays the same. When we encounter new errors in the serial protocol, we simply update the function with addition error handling. This approach allowed the most seamless integration between the GUI and serial data transfer.
On the Arduino, we first check to see if any data exists in the serial buffer. If we see data, we read 256 bytes from the serial buffer into our serial buffer array. We then update the HeatOn and MotorsOn variables with the first two bytes of the serial buffer array. The Arduino also sends current nozzle temperature and current filament diameter to the Raspberry Pi through the same serial line through the use of serial.Print() and serial.Println() functions.