Help Interpreting PID Controller Settings

[JustinAW]

I am somewhat familiar with a PID controller and how it operates. But I am having a hard time interpretting my PID controller setting and getting them to do what I want. I figured I would open up here for some help.

I have a copper cylinder that is being heated up with a band heater. in the center of the cylinder I have the thermocouple. Both are hooked up to my PID controller. Because of the lag due to conduction I would like my Temperature curve to approach the limit and slow down. More like an asymptote and less like a damping sine wave. My current setting get to the temperature fine but then way over shoot it and it takes a while to cool back down. This is not Ideal for the experiments I want to run with it.

Below are my settings options:

Proportional band setup :_____ in Celsius
Proportional Gain setup :_____ no units provided
Integral time setup :_____ in seconds
Integral Gain setup :_____ no units provided
Derivative rate setup :_____ in seconds
Derivative rate Gain setup:_____ no units provided
Power output control :_____ percent

 

[skeptic2]

Could you give us some more information about the controller - make and model number?

 

[Asymptotic]

Could you give us some more information about the controller - make and model number?

Also,
What are the cylinder dimensions?
What is the heater wattage rating?
A photo of the copper cylinder, heater, and thermocouple would help, too.

What type of heating output (time proportioned digital, or continuously variable analog)?

What are the current values for the settings you've listed. Are they the controller default values?

Does this controller have an auto-tuning function?
If yes, perform an auto-tune, and note what the setup values have changed to.

At first glance, it appears like your system has a lot of time lag.

 

[JustinAW]

It's a 2 inch diameter copper rod. The thermocouple is an E type. The controller is a digisense 9500 temp controller. I'm not sure about the heating output. I don't have a picture with me. Yeah its definitely a time lag. It takes a while for the heat to conduct to the center of the rod where the thermocouple is. When this info reaches the controller it has already heated up way past the goal.

I want to do some experiments to see how long it takes for the conduction to get through. That way I can adjust the setting so that it slows down way before the goal instead of after.

I just have no idea what I would adjust for the settings.

 

[JustinAW]

It does have an auto tuning function. I'll do that and see the default values later

 

[Bandit127]

Auto tuning should work. If you have to manually tune it, start by increasing the integral time constant. Use the magnitude of the overshoot for feedback and keep notes - it can be easy to lose your way when manual tuning.

The proportional band could be the temperature offset where the controller starts working proportionally. The ones I work with will heat at full current until they get to the offset and then control proportionally. (E.g. with a setpoint of 100° and an offset of 20° they will heat at full current to 80° and then control proportionally from there). Too small a setting here will create overshoot whatever the other settings are on.

 

[Asymptotic]

The documentation mentions USB-based communications and logging software, but not exactly what is logged. Measured value is a given, but it doesn't say whether it also logs loop output percentage and setpoint.

I've found it can be enlightening to closely monitor and log controller operation during auto-tuning, particularly the time lag between when output is turned on, and temperature rise is sensed (lag), and rate of temperature change once the curve straightens out. Read up on the Ziegler Nichols open loop (process reaction, aka step response) tuning method, and use trended data to double-check what parameters the auto-tuner comes up with. This overview may be helpful. http://blog.opticontrols.com/archives/477

I didn't find a time-proportioning cycle time variable in the manual. On controllers that I've used the default ranged from 20 to 60 seconds. Cycle time ought not affect your process negatively unless considerably longer than 30 seconds. It is easy enough to find out what the cycle time is - use a watch, and time how long it takes from output switching from off-to-on to the next cycle's off-to-on transition.

The controller does have an output limiter parameter (screen #17 - power output control) than can be cut back from 100% to an experimentally determined lower value to reduce overshoot. One downside of this approach is initial heat-up to setpoint takes longer.

To resolve a question that came up while scanning through the manual.

• Turn off the derivative term. It probably doesn't add much value here, but (especially if feedback is the least bit noisy) can aggravate instability.
  
• For the time being, turn off integration as well, and run as a pure proportioning controller.
  
• Note the proportional band and gain values in screen #17 and screen #18.
  
• Double the prop band value on screen #17, and see if it halves the prop gain value on screen #18.

Proportional gain is the reciprocal of bandwidth. I've never before seen a controller that had adjustments for both, and can't see how both terms can be independently variable. I and D parameters are similar, and both are shown in terms of gain, and of time. I'm used to seeing them parameterized in terms of one or the other, not both.