How valve failure modes interact with controller actions and process loads

Q: I found the following statement in an engineering manual: "In order to function at all, a feedback control loop must have a negative feedback, a 180° phase shift somewhere in the loop."
Can you please explain the meaning of this statement? And how does it impact the controller action (direct/reverse) of a feedback loop?

Bodhisatya
[email protected]

A1: Feedback loops operate the same way as most doctors do: by waiting until the patient gets sick, develops the symptoms of a sickness (error), and takes corrective action only after that. Since you asked only about that, I will not discuss feedforward control, which is similar to what Dr. Fauci is trying to practice.

All feedback control loops consist of three main components: 1) the process, 2) the control valve and 3) the controller (Figure 1). Each component can be either direct acting (DA), in which case its output changes in the same direction as its input, or reverse acting (RA) in which case the output responds to an input change by moving in the opposite, or reverse, direction.

A process is DA if a change in the process load (Q) results in a change in its controlled variable (C) in the same direction as Q. Conversely, if Q and C move in the opposite directions, the process is RA. For example, heating is a DA process and cooling is an RA process. Similarly, the control valve is DA if the flow (F) through it increases when the controller output M (manipulated variable) rises, and it's RA if F drops in response to a rise in M. For safety reasons, heating valves are usually DA (fail closed, FC), and cooling ones are RA (fail open, FO).

The controller action is selected for the total loop (Figure 1) to always have negative feedback (RA). Each RA (180° phase shift) in the loop causes a sign reversal, while each DA results in no phase shift. In terms of the total loop, this means that if both the valve and the process are DA (phase shift = 0) or if both are RA (phase shift -180° - 180° = -360°), the phase shifts cancel out. Therefore, for the controller to provide nevagive feedback, it must be RA. If on the other hand they differ (one DA with 0°, the other RA with 180°), for the loop to be RA, the controller must be DA.

It's the general practice for energy supply valves (steam, hot oil, etc) to fail closed (FC), and energy-removing valves (cold or chilled water) to fail open (FO). Here I'll make some general comments about the causes of valve failure and the selection of valve failure positions, but will only discuss the valve response when power supply to the actuator fails (loss of instrument air or electricity supply) because it would take too much space to discuss other failures, such as spring, diaphragm or piston failures.

When globe valves are used with pneumatic, spring-loaded actuators and direct-acting positioners or no positioners, the ultimate valve position will not only be a function of the actuator design, but also of process fluid forces acting on the valve itself. The valve design choices are FTO (flow to open), FTC (flow to close) or FB (friction bound) when the valve stays in it's last position (FL). FTO action is available with globe valves. FTC action can be obtained from butterfly, globe and conventional ball valves. Rotary plug, floating ball and segmented ball valves tend to be FB and the flow direction through them can possibly affect the torque required to open the valve.

Spring-loaded actuators are the most convenient means of providing FC or FO action, while two-directional air or electric motors will tend to fail in their last positions (FL).

Béla Lipták
[email protected].

A2: Integral-only control can work, which upends the 180° phase shift requirement.

Yes, a feedback controller seeking to keep or return a controlled variable to the setpoint needs negative feedback, as it needs to take control action to counter the deviation. But the question of direct or reverse action isn't so simple. Consider temperature control of hot and cold water mixing. If a hot water valve is used to control the temperature and the actuator is air-to-open (fail closed), then increasing the controller output opens the valve and increases the temperature. But, if the same valve is on the cold line and it's used for temperature control, then increasing controller output reduces the temperature. And, if it's a fail-open valve, the opposite is true again. Opposite again, that is, if overflow is used to adjust the process.

One must see all cause-and-effect mechanisms in the control system and process to determine which is right. The ISA definition of direct and reverse acting is based on regulatory mode; direct action means if the measurement increases, then controller output rises to fix it. Reverse action means if the measurement rises, then the controller output drops to fix it.

R. Russell Rhinehart
[email protected]

A3: Yes, this does impact proper setting of the controller action, direct or reverse. However, unless you're trained in engineering mathematics, don't let the words "phase shift" fool you. All you need to know is the relative direction of change of process element inputs and outputs. If, when the input to a process element increases, the output also increases, that is 0° phase shift, but let's call it something simpler: "no sign reversal." If on the other hand, when the input to a process element increases, the output decreases, that's a 180° phase shift, so let's call that a "sign reversal."

Now, let’s replace the statement that "there must be 180° phase shift around the loop" with the statement that "there must be an odd number of sign reversals around the loop.”

Examples:

1. Consider a normally open valve, which controls the flow of a heating medium to a heat exchanger. If the valve is normally open, a decrease in signal to the valve (i.e., controller output) will cause the valve to open. Consequently, the flow through the valve will increase. That's a sign reversal. When the flow of heating medium increases, the outlet temperature increases. That's no sign reversal. Since we already have one sign reversal, we don't need any at the controller. Therefore, we want an increase in PV (measured temp) to cause an increase in controller output. That is to say, we want error (difference between SP and PV) to be calculated Error = PV - SP. That's direct acting.
2. Consider a case that's the same as above, but the valve happens to be normally closed. That removes the sign reversal of the normally open valve in the example above. Since neither the valve nor the process provide a sign revesral, it's up to the controller. It must be set for reverse acting, or Error = SP - PV. An increase in PV will cause the controller output to decrease. That's the required sign reversal.
3. This case is similar to the process loop described above, except that the valve is normally open (that's one sign reveral). In this case, the purpose of the heat exchanger is to reduce the temperature of a hot incoming fluid, so an increase in flow rate of the cooling medium reduces the heat exchanger outlet temperature. That's another sign reversal. Consequently, we have two sign reverals, and we need an odd number. The next odd number is three, so the controller must provide the third sign reversal in the loop. In this example, the controller must be set for reverse acting.

In summary, to determine whether the controller should be set for DA or RA, think of all of the places where there will be an input-output sign reversal around the loop, and set the controller, so it provides an odd number of sign reversals.

Harold Wade
[email protected]