0
\$\begingroup\$

I have a HVAC pressure/flow controller that has a 0-10V output that is PID derived from the measured flow that the controller derives from two pressure ports.

What settings would I use to make the output track the measured flow, so that the controller is a flow sensor, not a flow controller?

The issue is that the error needs to be constantly zero (I guess the goal of all PID), but there is no system as such- the output will have no effect on the input.

I have tried online simulators, etc., but they all have the input a function of the output.

Has anyone met this for real, and how did it get resolved?

The context is we have 50 of these mounted on AHUs, a very tight timescale, and we need them to act as flow sensors.

Here is the idea based on Fred's comment.

Let the setpoint be zero.

Let X be the reading, I term and D term be as close to zero as possible (especially I otherwise we are well stuffed).

Then output Y = P x (error) = P(0-X) = -PX.

This negative value is not possible at the output, so set the setpoint to be the maximum flow expected, or F.

Then error is F-X, so output is P(F-X), or always positive.

P can be used to scale the signal, so that the output is of an adequate range. F is fixed by a setting on the controller.

It isn't brilliant, but it is better than rebuilding all the things with different parts.

I am hoping the control strategy only uses these as indicators, mind.

Update- the controller has P, I and D only, set as integers + decimal.

Update 2 - The anomaly with this is this is an open loop. The output has no effect on the input- there is no transfer function to link them together.

\$\endgroup\$
3
  • \$\begingroup\$ Assuming the setpoint as \$R(s)\$, output as \$Y(s)\$, PID controller + plant as \$H(s)\$ and feedback gain \$G(s)\$ (the sensor's transfer function), the expression for the output, with negative feedback, is: \$Y(s)=\dfrac{H(s)}{1+H(s)G(s)}R(s)\$. With setpoint equal to zero (\$R(s)=0\$), even nulling \$D\$ and \$I\$ gains, as suggested in an answers, the output will be zero. So, I think that such arrangement will not work here. It seems that the contributors are considering the loop as open and trying to track the setpoint and not the sensor. \$\endgroup\$ Commented May 28 at 1:53
  • \$\begingroup\$ Different PID controllers have different ways of adjusting their transfer function. Some will work for you, others won't. Please edit your question to say what you have. If you have P, I and D adjustments, just set D and I to zero. If you have a reset time constant, derivative time constant, and master gain, then unless you can set the reset time constant to zero, you're probably out of luck. \$\endgroup\$ Commented May 28 at 2:53
  • \$\begingroup\$ @REPuzzle : Now you have explicitly stated that the loop is open by including your Update 2. \$\endgroup\$ Commented May 28 at 16:46

2 Answers 2

Reset to default
1
\$\begingroup\$

At best a PID controller should produce the negative of the input signal if you can turn off the integral and derivative terms and set P=1. If there is an offset adjustment ('manual reset') you might be able to make it do what you suggest. Eg. bias it so that 'Zero' input is 'Full scale' output and vice-versa.

But typically controllers are not designed to do that and the output signal precision might be lacking.

Surely there's a better way. Sometimes there's a feature called 'retransmission' available, but not sure if that's a thing in HVAC.

\$\endgroup\$
2
  • 1
    \$\begingroup\$ Hello Spehro - yes, the problem with this seems to be a) nobody does it, and b) it is not the best way. I am looking into alternatives- exactly as you said- it is not intended to do that and so if someone comes along later they could blame any problem on the non-standard approach. \$\endgroup\$ Commented May 28 at 9:11
  • 1
    \$\begingroup\$ In the end we used another path to do it. The big issue was precision, or more precisely, proving precision to other stakeholders in the project. The controller in question could do Modbus, but if it did that, because of the cable it had, that connection would be a stub. In the end a repeater was fitted that split the Modbus into segments so now this connection was the end of the segment and no longer a stub. It was expensive, but much cheaper than changing the cable. It was also much easier as the Modbus allowed a direct export of the measured flow. \$\endgroup\$ Commented Jun 30 at 18:16
1
\$\begingroup\$

The question does not completely make sense but I will try.

In principle if you set the setpoint to zero, the proportional gain to a moderate value, and integral and derivative contributions to zero, then the output would be proportional to the input signal.

I have a feeling that any reasonable controller will not let you do this (I and D contributions will be finite but not zero).

What you need to do is get access to the flow measurement signal. End of story. Get the control action out of the picture. Again, it may not be practical to do this.

\$\endgroup\$
2
  • \$\begingroup\$ Hello Fred, that is pretty much the way I was thinking. I've added a comment to my post that details how it would be set to scale correctly. I would much rather access the signal but it it not an option in this case. \$\endgroup\$ Commented May 27 at 21:29
  • \$\begingroup\$ In the end we did go down the route of getting to the signal another way. \$\endgroup\$ Commented Jun 30 at 18:12

Your Answer

By clicking “Post Your Answer”, you agree to our terms of service and acknowledge you have read our privacy policy.

Start asking to get answers

Find the answer to your question by asking.

Ask question

Explore related questions

See similar questions with these tags.