Confessions of a Looney Tuner – "See no Evil, Hear no Evil, Speak No Evil"

By Sylvain Millette
Millette Control Engineering Inc.
millettecontrol@netscape.ca

I firmly believe that, given a set of initial conditions (and enough time), everything will gravitate towards their natural state. In this month’s adventure, I was again giving a loop tuning/troubleshooting lecture in the field. In this case, my students and I were hooking up to the input and output signals of a control loop. The loop was in the stock preparation area, a stock flow controller feeding a screen. A screen is designed to remove particles from the pulp stock based on size. Large flow and consistency variations can cause the screen to plug up. The only way to unplug the screen is to shut down that part of the process – a costly and disruptive operation. In this case, the flow transmitter was a magnetic flow tube type, and the valve was a six-inch, "V-Ball ".

Once hooked up to the flow control loop, we proceeded to perform bump tests on the loop. That is when things turned very interesting: every time the output was bumped, the flow responded in a "smooth" fashion, taking 60 seconds to settle out. This response was so "clean", so "smooth" and perfect that even theory would be challenged to reproduce such a response. I knew that nature itself would not be so accommodating, so something was afoot.

I asked my class whether it made sense that a stock flow would take 60 seconds to settle out, given the fact that the valve was only 6 inches. They all responded "No", it did not make any sense. I then asked them what could be causing such a slow response, one answered "transmitter damping" (the linear thinker) and another answered "the valve" (the lateral thinker).

We decided to test the second theory first. With the advent of smart valve positioners, it is possible to change, and slow down the positioner tuning. This in turn would slow down the response of the valve. Generally, I discourage people from adjusting these parameters since they are too hidden. We had one student perform bump tests from the control room while the rest of us stayed and "felt" the valve stem. These "finger tests" on the valve showed "crisp" movements to output changes. Clearly, the valve was not part of the problem.

The next step was to check the damping setting in the transmitter. We knew that the damping had to be in the transmitter since we were measuring the signal directly from the transmitter itself. Damping is typically applied in transmitters or controllers in order to remove noise from the signal. People seem to have an aversion to noise and prefer to see straight lines on their process trends. After some lengthy navigation through the menus in the flow transmitter, we finally arrived at the transmitter damping setting: it was set at fifteen seconds. I typically consider two seconds more than enough for a flow, let alone fifteen seconds! Such a large amount of damping not only masks any noise in the signal, it also hides the actual process dynamics. But the question remained: why would someone put such a large amount of damping in the transmitter? What was their reasoning (for everything happens for a reason)? "Obviously, there must be a lot of noise on the signal" one of my student said. I then asked a more fundamental question: why was there excessive noise on the flow signal? No one could immediately answer that question. We proceeded to go out and physically inspect the installation of the flow tube itself.

Upon examining the transmitter installation, our question was promptly answered. The flow-tube was flanged to a piping elbow; also less than 12 inches downstream of the tube was the control valve. Magnetic flow tube flow transmitters require a stable plug flow in order to develop a stable and clean signal. Any source of turbulence such as elbows, tees and valves will affect the flow stability and cause signal noise. In this case, every possible rule for installing a magnetic flow transmitter was broken. Therefore the solution was to put excessive damping in the transmitter in order to cover up for a mechanical process design fault.

See no evil, hear no evil, speak no evil…