Editor's Choice


Loop signatures 20: The great deceivers.

October 2023 Editor's Choice

Noise on the PV signal was discussed in the previous loop signature article, and it was mentioned that it is generally better to live with the noise rather than filtering it out, provided it does not cause the final control element to jump around excessively. This article explains this reasoning.

Normal filters used in process control have a first order lag type dynamic, which is generally preferable to time averaging filters. This is because PID controllers can handle lag dynamics better. A lag type filter is also referred to as a low pass filter, as it passes the lower frequencies and attenuates higher frequency components. Visually, the filtered signal appears to have less noise. Obviously, the greater the filter constant, the smoother the output signal.

In the old analogue days, filters were extremely difficult to implement, and as a result were seldom used. Today, it is extremely easy to implement a simple lag type filter on a digital control machine. Only a few lines of code are needed. Most DCSs and a few PLCs offer filters as a standard block. These filters usually allow one to insert lags in a loop ranging from fractions of a second to literally hours.

When a filter is used in the control equipment, it is usually inserted before the controller and the operator’s display, or the scada. Figure 1 shows the response of a flow signal to a step change on the PD (controller output). The signal has noise of some 2% amplitude on it. The signal is then passed through a filter with a five second time constant, and the output of the filter is shown in the same figure superimposed over the original PV signal. It can be seen that the noise amplitude on the filtered output has been reduced considerably. There are still some fluctuations, due to the lower frequency components present. Apart from that, the signal has not been changed drastically.

In Figure 2 one can see how the situation has changed when a 50 second filter is used. The noise has now been completely eliminated. The filter has certainly achieved its purpose inasmuch as it got rid of the noise. However, the signal response has been really drastically slowed down. The filter has now replaced the flow process as the dominant dynamics in the control loop. In other words, we are now trying to control a filter rather than a flow.

Many learned texts on control have stated that if the time constant of the filter exceeds the dominant time constant of a self-regulating process, the control will be adversely affected. This is not necessarily correct, as you will see in the next example. Tests we have performed show that you can in fact get quite reasonable control on loops with relatively heavy filters. However, there are other really serious effects which are more important, and which make the use of big filters very undesirable.

One of these seriously bad effects is that filters can cause excessive PV overshoots (as high as 300%) on step-changes in setpoint. Such an overshoot is shown in the next example. In previous articles, I have also referred to filters as the world’s biggest liars, as they can distort the response of the process, as seen on scada or DCS screens by the operator. The filter actually tends to reverse the facts, making good things look bad, and bad things look good.

To illustrate this point, Figure 3 is the recording of a control system on a noisy process. The controller has a filter with a 50 second time constant on it. The controller is extremely well tuned. The recording is made with the Protuner analyser connected at the input to the controller (upstream of the filter). A 10% step change in setpoint is followed by a 10% step change in load. An enormous overshoot on the PV response to the setpoint change can be seen. It is really bad. The load change however is well controlled. The controller brings the process back to setpoint speedily, with little overshoot. This is in fact really good. It also shows you can indeed get quite good control, even with a big filter.

Figure 4 shows what the operator sees on his screens with a filtered signal. The setpoint change looks good. There is no sign at all of an overshoot. However, the load change now looks terrible. It only shows as a tiny offset, which seems to take forever to get back to setpoint. Therefore the filter has completely distorted and reversed the true facts. This to me is the most dangerous thing of all, as the operator does not really know what is going on in the plant. It must be remembered that the people doing the tuning are generally also unaware of what filters can do, so they often ruthlessly eliminate noise with enormous filters, and then do tuning whilst looking at the filtered signal, so they have no real idea of the true response.

It is common practice these days for management to use the information available on the scada for management evaluation and control purposes. Therefore it is vital that people should avoid using big filters if they want to have reliable information. This was illustrated on a mine in southern Africa where all the plant accounting and management information was based on a transfer from the scada package into spreadsheets. Large filters were applied on nearly every loop. It was a continuous source of puzzlement to the metallurgists as to why the plant balances never seemed to work out.

Rules for filtering

Sometimes one has to apply filters when the noise is too much and starts moving the final control element around too much. If you apply the following rules of thumb you will find that you cannot go too far wrong.

• If possible, eliminate the noise at its source.

• Use an anti-alias filter on the transmitter (to be discussed in a future loop signature).

• Try tuning without a filter, and check at the actual valve if it is moving due to the noise.

• Avoid using a bigger filter than necessary.

• Biggest filter should be less than 10 x (process deadtime + scan rate of controller).

• Use I-PD controller with filters (see loop signature 16).

• Train the operators and process people on filter effects.

The last point is one of the most important. Unfortunately process personnel are so used to heavily filtered signals in most plants that they are very unhappy when filters are removed. As mentioned in the last loop signature, they mistake noise for instability. It is sometimes very hard to convince them that the control is now much better.

It was mentioned earlier that if you do have to use a filter, then it should be inserted in the control equipment rather than the transmitter. This is because you cannot see the control response at all if you filter in the field. The one exception to this rule is an anti-alias filter, which will be discussed in a future loop signature.


About Michael Brown


Michael Brown.

Michael Brown is a specialist in control loop optimisation, with many years of experience in process control instrumentation. His main activities are consulting and teaching practical control loop analysis and optimisation. He now presents courses and performs optimisation over the internet. His work has taken him to plants all over South Africa and also to other countries. He can be contacted at: Michael Brown Control Engineering CC, +27 82 440 7790.


Credit(s)



Share this article:
Share via emailShare via LinkedInPrint this page

Further reading:

STEMulator – a gift to the youth of the nation
Editor's Choice News
STEMulator is a groundbreaking virtual platform designed to ignite the spark of curiosity in young minds and stimulate their interest in STEM subjects.

Read more...
Innovate, accelerate, dominate
Festo South Africa Editor's Choice Pneumatics & Hydraulics
Festo’s latest innovations, revealed through the Ramp Up Campaign, offer a blueprint for performance excellence, using the anatomy of a race car as an analogy to simplify and powerfully communicate how their technologies address industry challenges.

Read more...
Case History 198: Cascade control overcomes valve problems.
Editor's Choice Flow Measurement & Control
There are many processes where it is undesirable for the load to suddenly change quickly, for example in the paper industry. Examples of level control have involved reasonably fast tuning. An example of a level loop tuned this way and responding to a step change in setpoint is given.

Read more...
Advanced telemetry solutions
Editor's Choice Industrial Wireless
Namibia is one of the driest countries in sub-Saharan Africa, with an average annual rainfall below 250 mm. To address this challenge, the Namibia Water Corporation has employed one of southern Africa’s most powerful and well-proven telemetry solutions, designed and manufactured by SSE/Interlynx-SA.

Read more...
Navigating the future of intralogistics
LAPP Southern Africa Editor's Choice
In the rapidly evolving landscape of global markets, the demand for agility, efficiency and scalability in intralogistics has never been more critical. At LAPP Southern Africa, we stand at the forefront of this transformation, offering cutting-edge connection solutions tailored to the dynamic needs of intralogistics.

Read more...
Cutting-edge robotics and smart manufacturing solutions
Yaskawa Southern Africa Editor's Choice
Yaskawa Southern Africa made a compelling impact at this year’s Africa Automation and Technology Fair.

Read more...
A cure for measurement headaches in contract manufacturing
VEGA Controls SA Editor's Choice
A contract manufacturing organisation provides support to pharmaceutical and biotechnology companies in the manufacturing of medications, formulations and substances. VEGA’s measurement solutions offer accuracy and reliability for monitoring levels and pressures during the manufacturing process.

Read more...
PC-based control for a food capsule and pod packaging machine
Beckhoff Automation Editor's Choice
For TME, a machine builder specialising in the packaging of powdered foods, Beckhoff’s PC-based control technology offers unlimited opportunities when it comes to performance and innovative capacity in terms of flexibility, scalability and openness.

Read more...
Case History 198: Cascade control overcomes valve problems
Michael Brown Control Engineering Fieldbus & Industrial Networking
A large petrochemical refinery asked me to perform an audit on several critical base layer control loops. This article deals with a problem found on a valve controlling the flow of fuel to a heat exchanger.

Read more...
Simple and efficient level measurement in the mining, minerals and metals industries
Endress+Hauser South Africa Editor's Choice Level Measurement & Control
Measuring devices in the mining, minerals and metals industries face the challenge of varying material states and long distances in measurement height. Endress+Hauser’s answer to these challenges is the new Micropilot family.

Read more...









While every effort has been made to ensure the accuracy of the information contained herein, the publisher and its agents cannot be held responsible for any errors contained, or any loss incurred as a result. Articles published do not necessarily reflect the views of the publishers. The editor reserves the right to alter or cut copy. Articles submitted are deemed to have been cleared for publication. Advertisements and company contact details are published as provided by the advertiser. Technews Publishing (Pty) Ltd cannot be held responsible for the accuracy or veracity of supplied material.




© Technews Publishing (Pty) Ltd | All Rights Reserved