1. Home
  2. ⧑ Industrial Control Group
  3. Re: Odd step response

Re: Odd step response

Dave,

Nor do operators like odd looking time domain responses. In a real, non-linear plant, only the time domain matters in the end. The bottom line is, "How does the plant behave over time?"

1 - Are any parameters outside safe boundaries, i.e. maximum over shoot too great?

2 - Is the blended product composition off spec, i.e. average error too great?

3 - Is any of the equipment being overstressed, i.e. rate of change of output, or PV, too great?

4 - Does the PV return to setpoint quickly enough that the resulting error can be averaged out in the tankage?

5 - Is the process stable across the entire range of operating conditions, i.e. closed loop gain < 1 despite uncontrollable changes in process gain?

You will note that none of these parameters are readily identifiable in the frequency domain but they sure do show up in a time domain plot.

Frequency domain makes entertaining math. Time domain makes money.

Walter.

> Does anyone have some insight into the causes of this behaviour and how > to eliminate it. I tend to tune in the frequency domain, but our system > validation testers prefer to check in time-domain and don't like 'odd- > looking' time responses even when they meet the frequency-domain > specifications. (Nor do I).
Reply to
Walter Driedger
Loading thread data ...

What is the plant?

This is pole placement. Sometime this can result in some funny ( negative ) PID gains or in your case lead poles and zeros and gains.

I do a lot of hydrauliuc pressure control. Your description matches what the closed loop pressure responses does when the actuator hits and obstruction and the differentiator gain too high. Your tuning is a little aggressive, but it would probably work fine IF you ramped your SP instead of stepping it.

This is a resonance peak often found in systems with imaginary poles.

It is the unmodelled in a system that makes this a challenge.

Peter Nachtwey

Reply to
Peter Nachtwey

That's got to be the best "Controls" quote I have heard thus far. Can i borrow it ?

-kamesh

Reply to
Loony Tunes

In article , Loony Tunes writes

In article , Loony Tunes writes

I like it too ! And that's a great set of check-points Walter gives.

However, the 'odd' response would pass many time-domain requirements. Overshoot is fine (it doesn't), rise time is fine (rises quick at first, then drops then slowly rises). The plant is not overstressed. The settling time, however is probably unacceptable. The main problem is that the plant will be driven by a human operator on the throttle, who, I suspect, would find the response 'odd' in the same way I do.

So I offer this caveat to the quote.

Frequency domain is USEFUL entertaining math for computer-aided design. I can go point the spec at my computer and let it go calculate the gains, at all 1000 operating points while I go do something useful like drink coffee or possibly work out what my customers REALLY meant by those requirements... Out of these 1000 points, about 5% seem to come out with responses which I would deem odd, but the algorithm wouldn't spot. At this rate (which would depend on the system in question) I would reckon that frequency-domain design is cost-effective. I can always go tune the odd cases by hand.

But, I'd agree, time-domain validation is absolutely essential.

regards

Reply to
Dave Hill

PolyTech Forum website is not affiliated with any of the manufacturers or service providers discussed here. All logos and trade names are the property of their respective owners.