industrial//instrumentation//control valve//stiction

Stiction is the static friction that holds a mechanism still until the force on it exceeds a threshold, after which it breaks free and jumps, and in a control valve it is the most common cause of a loop that oscillates by itself however it is tuned; the oscillation, called hunting, wastes energy and product and wears the valve. The source is **dry friction**, or Coulomb friction, between the stem and its packing: the force needed to start the stem moving is larger than the force needed to keep it moving, so the valve sticks, then overshoots the position it was asked for.

Stiction is the static friction that holds a mechanism still until the force on it exceeds a threshold, after which it breaks free and jumps, and in a control valve it is the most common cause of a loop that oscillates by itself however it is tuned; the oscillation, called hunting, wastes energy and product and wears the valve. The source is dry friction, or Coulomb friction, between the stem and its packing: the force needed to start the stem moving is larger than the force needed to keep it moving, so the valve sticks, then overshoots the position it was asked for.

The loop turns that jump into a cycle. The controller asks for a small change, the stem does not move, the error persists, and the integral term ramps the command up. At some point the force exceeds the static friction, the valve jumps past the position the process needed, the error changes sign, and the integral ramps back the other way until the valve jumps again. The controller output traces a sawtooth while the flow traces something closer to a square wave, a limit cycle of fixed amplitude that persists whatever the starting point. Lowering the gains slows the cycle down and does not remove it, because the cause is a nonlinearity in the actuator, which no linear tuning reaches.

It is diagnosed with the controller in manual. Moving the output in small steps and watching the stem or the flow shows the dead band directly: nothing happens for the first steps, then a jump. In routine data, plotting the controller output against the flow draws a parallelogram where a healthy valve draws a line, and plant audit tools look for that shape.

The fixes are mechanical first. Repacking or replacing the valve, a positioner tuned to break through the friction, and a smaller integral action (or none, on that loop) reduce the cycle; the lasting cure is maintenance, since retuning only changes the cycle's period and amplitude (control valve).

The same friction lives in every mechanism. Robot joints, linear stages and gearboxes show it next to backlash and dead zones, the ordinary nonlinearities that sit between a command and the motion it produces (actuator).

Its signature misleads fault-finding. A loop that oscillates with a constant period looks badly tuned, and many loops are retuned, softened or put in manual for a fault that lives in the valve; checking the actuator before the gains is the first item of PID tuning.