OPT FOR SECURITY AND PERFORMANCE

Improve the security and performance of your plant with alarm management, performance supervision, asset supervision, web-based data visualisation, optimisation, advanced control, and tuning with the new TaiJi tool.

Rethink your working methods and obtain immediate results thanks to our services, trainings, and consulting expertise. Manage changes intelligently with our proven solutions.

 certified service provider

November 2006

In this issue

puce Classes
puce Trick of the month: Origin of oscillations
puce To contact us



Coming classes and seminars

Seminar on modern control methods
November 29-December 1, Bangkok, Thailand and
December 5-7, Malaysia

Control Valve Selection
November 14-16, Moncton, NB

PID Loop Tuning
November 21, Ottawa, ON

Introduction to Instrumentation and Process Control, Process Control Basics for Non-Specialists
November 22-23, Ottawa, ON

See our calendar for 2006

This month's tip : Origin of oscillations

You observe oscillations in temperature measurement (PV) and ask where are they coming from?

The root cause of an oscillating loop may not be obvious to determine. There are many possible factors that can contribute to the oscillation. First, we have to determine whether the oscillation is related to the tuning, the hardware or the disturbance (load). If it is possible to distinguish a sinusoidal pattern on a trend during a steady-state period of time (during a constant load), then it is likely to be correlated to the tuning. These oscillations could be coming from other loops or other processes. Put the loop in manual mode: if oscillations cease, it means that these oscillations were induced by the controller itself. However, four parameters are involved: can we quickly pin point which one is the problem? Many controllers only use PI, hence only three parameters remain: Proportional, Integral and Filter.

Because either Proportional or Integral have equal potential to make the loop oscillate, it would be interesting to find a technique to investigate which of these two is the poor performer. To do so requires a minimum knowledge of the dynamic: this is the period of oscillation and dead time.

The dead time is the time that elapses after a controller output change until we observe movement in measurement. The controller output change can originate from a set point change or from a manual output variation.

Symbols:

Period of oscillation: tosc (s)
Dead time: td (s)
Time constant: (s)
Filter time constant: (s)
Integral time: TI (s) (if controller structure is parallel, convert to ideal or series)
Derivative time: TD (s) (if controller structure is parallel, convert to ideal or series)
Proportional gain: Kp

Filter

If < 4* td, ignore the filter time constant: the filter has no impact on the period of oscillation. Otherwise, temporarily remove the filter and measure again tosc. The filter could be in the transmitter, analog input card, filter function or controller function.

Derivative

If TD < ½ * td, ignore it: derivative has no impact on the period of oscillation.

Proportional or Integral

We distinguish two classes of processes:

  • Dominant dead time, td
    • Flow loops, Consistency, Conveyor weight or flow, etc.
  • Non dominant dead time, td <<
    • Temperature, Level, Pressure, etc.


Dominant dead time, td Non dominant dead time, td <<
If tosc/TI < 2, then P is guilty If tosc/TI < 1, then P is guilty
If tosc/TI > 4, then I is guilty If tosc/TI > 2, then I is guilty

Conclusions

Long oscillations (compare to Integral time, series or Ideal algorithm) usually originate from too much Integral action.

Short oscillations usually originate from too much Proportional action.

Very short oscillations could originate from too much Derivative action.

Reminder for series or Ideal algorithms:

< TD < td < TI < tosc

To contact us

puce Lévis's office +1 (418)834-4321
puce Pointe-Claire's office +1 (514)695-3492
puce Hartland's office (877)867-6473
puce Toll free (877)867-6473
puce Email info@topcontrol.com