Hi all,
I'm new in the forum. I would like to know if there is a way to estimate the total accuracy of a control system composed by a digital PID that regulates a pump with a feedback on a flow sensor.
Considering the accuracy of the sensor component, what are the other variables to take into consideration to calculate the total accuracy? Is it possible to estimate in a theoretical way all the errors?
Thanks

 

The accuracy of the control is in my opinion how many % the measured value deviates from the setpoint, and how many deviations occur during actual control of the process per time unit.
What is also very important, is how fast the controller reaches its setpoint from startup or when the setvalue changes. And how much overshoot/undershoot occurs when the new setpoint is being reached.

If you take a very good temperature controller, with excellent control algorithms, the accuracy and responsiveness of the control will be excellent.
See for example our Japanese RKC Instrument GZ series, 100 measurements & PID calculations per second and pro-active overshoot and undershoot suppression techniques.
With this excellent GZ controller we heat a silicon wafer to 1200°C in just a few seconds without overshoot.
The sensor is also very important, it should measure the process/product temperature as accurate & fast as possible. We have for example ultra fine thin 50um (micrometer) type K thermocouples with response times of just 0.03s from ambient to 250°C.
See: https://cascade.net/en/gz-series/

 

If you take a very good temperature controller, with excellent control algorithms, the accuracy and responsiveness of the control will be excellent.
See for example our Japanese RKC Instrument GZ series, 100 measurements & PID calculations per second and pro-active overshoot and undershoot suppression techniques.
With this excellent GZ controller we heat a silicon wafer to 1200°C in just a few seconds without overshoot.
The sensor is also very important, it should measure the process/product temperature as accurate & fast as possible. We have for example ultra fine thin 50um (micrometer) type K thermocouples with response times of just 0.03s from ambient to 250°C.
See: https://cascade.net/en/gz-series/

Are you suggesting using a temperature controller on a pump flow application? That would be interesting to see.

What the OP needs to know is how his system works and he has provided us with little information about the pump and piping.
There is a system curve and a pump curve. The intersection of the two curves determines the pressure and flow operating point. This is not a linear system unless the pump is a fixed displacement pump but we don't know that.
The answer to the OP's last question is yes.

 

No the temperature was just an example.
With a fast response controller like the GZ you can also easily control pressure or flow. The principle remains the same as with temperature: measure process conditions (pressure & flow) as fast & correct as possible.
Non-linearity is also that is solved in our GZ controller, you can correct the input signals for that, so also no problem.

 

No the temperature was just an example.
With a fast response controller like the GZ you can also easily control pressure or flow.

Really? Do you really think 100ms or even 10ms updates is fast enough? It might be depending on this application but I know of many applications where it isn't. I think understanding the problem is most important. Then you can pick a control method.

The principle remains the same as with temperature: measure process conditions (pressure & flow) as fast & correct as possible.

So all systems are created equal? Just throw a fast PID at it?

Non-linearity is also that is solved in our GZ controller, you can correct the input signals for that, so also no problem.

It is a problem if you don't understand the problem.

Fast PIDs are a dime a dozen.
So how does your controller solve the OP's question on estimating the accuracy of his system?
I sell motion controller with updates much faster than than your controller. My controller can not solve the OP's problem.
This isn't about controllers. It is about understanding the system. Can you model the motor? Can you figure out there the operating point is? What if the system curve changes due to a valve being opened or closed down stream. Can you model the response? Controllers are just a tool. Knowledge is the key and that is what the OP is asking for. So the answer to the OP's last question is yes, the system can be modeled and accuracy estimated. However, it will take some work and controllers can't do that.