PROCESS VARIABILITY and control valve tests

A new process variability testing method helps to evaluate performance of a control valve BY Jari Riihilahti and Jouni Pyotsia

Control valves are often the most critical components in control loops, and their performance has a strong influence on process variability. This performance has to be known in optimising the control loop performance.

Over-sized valves, often due to poor process data with mixed phase fluids, and lack of valve maintenance, are the most common reasons for poor control performance. The solution is a new process variability testing system, which can also be evaluated in advance.

In the system, a real control valve is connected to a control loop simulator, with simulation models selected from a model library. The system helps understanding of dead band, dead time, hysteresis, speed of response, repeatability and linearity.

To recap, control performance depends on the installed flow characteristics and gain, as well as static and dynamic behaviour of the valve.

The best performance is usually achieved by a double acting actuator and a two-stage electropneumatic positioner. Double acting actuators are less sensitive to disturbances, like dynamic torque or force, and they also give a fast valve response. It is also important that the I/P conversion of an electropneumatic positioner is inside the position feedback, to compensate for feedback mechanism errors.

Characteristics and gain

If a valve is over-sized, only a small part of the control range is in use. Many process controllers still have 10 bit D to A converters, which means about 0.1% resolution for the control signal. The theoretical resolution of the control is 0.5% if only one fifth of the full control range is used.

Over-sizing means poor control accuracy. For example, lets consider the situation in which the absolute position accuracy for valve travel is 0.5%, and the maximum operating point is about 40% of the full travel. In this case, the control valve maximum relative installed gain is about four in a typical liquid flow application. This means that the maximum accuracy in flow rate terms is 2.0%.

The control valve installed flow characteristics and gain also affect strongly control valve dynamics and dead time. If the gain varies strongly within the process operating range, then the operating speed (dead time and time constant) of the control loop varies strongly. Control valve over-sizing increases control loop dead time.

Gain of an over-sized valve is high, such that the controller gain must be reduced to prevent control loop instability. This means that the control valve input signal changes, coming from the process controller, are smaller, which leads to a strong increase in control valve dead time.

Behaviour characteristics

Static behaviour of a control valve can be described with many different factors. Dead band, hysteresis plus dead band, repeatability and conformity/linearity are all defined, for example, in the standard ANSI/ISA S51.1-1979 (2).

The first two are the most commonly measured non-linearities of a control valve. They are mainly caused by backlash and high stiction in the valve. Backlash can be removed only by valve maintenance. The actuator load factor can be decreased by reducing static friction, and selecting a strong enough actuator.

Dynamic performance of a control valve can be defined by dead time, time constant(s), overshooting, settling time, bandwidth and stick-slip. The first four factors can be measured from a step response curve. Results depend greatly on the step size and the valve initial position, because the dynamics of a control valve are not linear.

A new testing system is now available. Process variability testing is a tool for evaluating the loop performance without a real process. The system is based on a real control valve, data acquisition system and a control loop simulator. The system simulates a simple unit process, process controller, process transmitter and disturbances. The flow characteristic of the valve is also modelled in the system.

Different models

There are different unit process models, such as a tank level loop and a flow loop. An operator can compare different process controllers and tunings against valve response. Also, signal disturbances and noise can be set by the operator.

Conventional tests, such as dead band and speed of response, are not adequate in forecasting performance of an installed control valve. The main benefit of this testing system is its simplicity, as it estimates control valve performance on process variability, using a simple graphical interface.

* The Author is with Neles Jamesbury. This feature is based on a paper given at ISA 96 – see Books, C&I, Dec page 38.