Are these sorts of influences noise or real process changes? Something caused the measurement to have perturbations. The perturbation is caused by something real, but that something might not mean the process changed. For example, the process flow rate might be steady (unchanged), but turbulence or vibrations made the measured flow rate value change. Or, in an inline blending, the mixing might be incomplete, and a composition or temperature measurement may vary as rich, or poor packets pass by the sensor.

However, the noise might represent small and temporary process changes from an uncontrolled influence.

Is the perturbation “noise”? From a data-based view, it depends on the sampling frequency of the observation. Nois” is uncorrelated. Sampling faster, while the influence of the something real persists, will make the perturbations correlated. Making it appear that the process is drifting.

From a process control view, it also depends on the speed with which the control system can correct the deviation. The deviation might reveal a real process change, but because of transport delay or analyzer delay, or batch processing, the material may have passed through the process and controller will not be able to correct it.

If the effect passes before the controller can correct it, or if the perturbation is noise rather than a real persisting deviation, then a controller would simply be tampering—responding to noise that it can’t fix. The cause of the perturbation will be gone before the controller correction is implemented, and the controller will just be amplifying the noise by adding another source of deviations.

Don’t try to control noise. The speed of the control system to correct a PV upset, must be much faster than the duration of the upset.

Ensure instruments are calibrated

Certainly, an improperly calibrated sensor/transmitter can introduce a systematic error. For example, time, heat, degradation of internals, or erosion may have made a once-calibrated device to now be out of calibration. Additionally, if the sensor to transmitter response is nonlinear (for example, temperature to thermocouple mV), yet it was calibrated assuming to be linear with a two-point calibration, then between the two calibrated points the sensor will report a systematic error. Nearly all sensors are nonlinear, but within a reasonable range, the linear assumption is acceptable for most.

Be sure that instruments are still in calibration and that calibration procedures match the sensor-to-measurement reality.

Match the solution to the character of the noise

Noise solutions include first-order filtering, statistical filtering, median filtering, improved resolution, nonlinear compensation and less frequent sampling. The solution must match the noise character.

Detecting truth through the noise

As part of a prior discussion about data, Ed Farmer suggests an even broader context, “Sometimes, though, it becomes essential to differentiate the underlying reality from the corrupted measurement data.”

Get your subscription to Control's tri-weekly newsletter.

Filtering tempers noise, but adds a lag or delay, and does not correct measurement bias.

To detect reality, seek to minimize sources of the various types of perturbations or misrepresentations in the measurements. Keep instruments calibrated and consider data reconciliation to correct measurements that have a systematic bias.

Measurement bias might change on a daily or weekly basis due to sensor device damage, progressive failure or changes in operating conditions. Data reconciliation adjusts measured values to make material and energy balances about process sections close. It uses variance on each measurement source, to determine uncertainty.

Human communication

Filtering may be desirable to temper control action response to noise, but observing the filtered variable masks features that an operator may want to see, such as when a transient starts or stops, whether there are spurious signals, etc.

It may be beneficial to let operators see the noisy signal on their display, rather than the filtered signal which is used for control or online analysis.

Noise is not just a part of process control systems. Erroneous messages and noise also refer to distraction from or degradation of a message, background nonsense or confusion, such as which is promulgated by social media, distortionists, or propagandists.

Humans lie to cover up or distract attention from things they don’t want to make known. Humans preserve traditions as reality, even when the tradition is technical folklore, perhaps a legacy misunderstanding of process mechanisms or a feature of the process that no longer exists.  Humans claim certainty whether they are certain or not, because uncertainty does not engender followership action or their future leadership-based promotion.

Use your understanding of filters to reject human-generated nonsense. Median filters to reject extremes, fact checking, mechanistic cause-and-effect reasoning or data reconciliation.