PID controller and possibles way to improve it
Posted by bravach
|
PID controller and possibles way to improve it March 29, 2013 03:46AM |
Hi folks,
When printing yesterday, I was struck by the fact that the way the temperature is controlled on our printer is not the way to go for fast speed printing. Indeed, the PID controller is a feedback loop which means that it only begins to raise power when he sees that the temperature is droping (and, in a similar way, lower the power when temperature is raising again).
Also, for the thermistor to see the change in temperature, it requires a lot of times from the moment the temperature changed at the nozzle.
This is purely logical and driven by the transfer function of the system behaviour but you can see it clearly on Pronterface that the temperature will drop seconds after a layer has been infilled at high feed rate.
In the industry, we don't use this kind of controller because it means that the extrusion temperature is not constant with the feed rate and can lead to a product out of its allowed range of quality. Actually, we use a combined regulator based on a feedforward loop to ensure that the proper power is applied before the extrusion begins and a smaller feedback loop to compensate the error of modelization.
It means that if you can measure the perturbation on the system, you can overcome it before they impacts your controlled variable or, we know the perturbation before they appears since the feedrate is in the G-code. It should be fairly simple to read some lines in advance (like firmwares do with SD printing mode) and calculate the proper PWM duty cycle to bring the polymer to its required temperature.
To do that, we must have the filament temperature before the extruder (I always have an additional thermistor hanging in the air to measure room temperature), the required temperature, the polymer specific heat, the hotend power at 100% duty cycle, the hotend thermal dissipation at rest (which also depends on the required temperature) and the amount of plastic to be fed.
Alas, PLA doesn't have a constant specific heat and it requires a look-up table to rapidly calculate the proper amount of energy required. Also, it would means experimenting with each sample of filament to get this look-up table. But, it should be possible to automatize this procedure (like the autotune PID feature).
I don't have much time right now to experiment with that but I did some simulations using MATLAB/Simulink as a proof of concept. I know that it is not practical for average people to do that but it could be an interesting proof of concept for higher feedrate printing if someone doesn't know how to spend his free time !
Also, for the record, I'm an industrial engineer in automation, so I know what I'm talking about.
When printing yesterday, I was struck by the fact that the way the temperature is controlled on our printer is not the way to go for fast speed printing. Indeed, the PID controller is a feedback loop which means that it only begins to raise power when he sees that the temperature is droping (and, in a similar way, lower the power when temperature is raising again).
Also, for the thermistor to see the change in temperature, it requires a lot of times from the moment the temperature changed at the nozzle.
This is purely logical and driven by the transfer function of the system behaviour but you can see it clearly on Pronterface that the temperature will drop seconds after a layer has been infilled at high feed rate.
In the industry, we don't use this kind of controller because it means that the extrusion temperature is not constant with the feed rate and can lead to a product out of its allowed range of quality. Actually, we use a combined regulator based on a feedforward loop to ensure that the proper power is applied before the extrusion begins and a smaller feedback loop to compensate the error of modelization.
It means that if you can measure the perturbation on the system, you can overcome it before they impacts your controlled variable or, we know the perturbation before they appears since the feedrate is in the G-code. It should be fairly simple to read some lines in advance (like firmwares do with SD printing mode) and calculate the proper PWM duty cycle to bring the polymer to its required temperature.
To do that, we must have the filament temperature before the extruder (I always have an additional thermistor hanging in the air to measure room temperature), the required temperature, the polymer specific heat, the hotend power at 100% duty cycle, the hotend thermal dissipation at rest (which also depends on the required temperature) and the amount of plastic to be fed.
Alas, PLA doesn't have a constant specific heat and it requires a look-up table to rapidly calculate the proper amount of energy required. Also, it would means experimenting with each sample of filament to get this look-up table. But, it should be possible to automatize this procedure (like the autotune PID feature).
I don't have much time right now to experiment with that but I did some simulations using MATLAB/Simulink as a proof of concept. I know that it is not practical for average people to do that but it could be an interesting proof of concept for higher feedrate printing if someone doesn't know how to spend his free time !
Also, for the record, I'm an industrial engineer in automation, so I know what I'm talking about.
|
Re: PID controller and possibles way to improve it March 31, 2013 03:39PM |
Alas, I'm using Marlin which doesn't support this feature. Thank you for the input anyway.
Sorry, only registered users may post in this forum.