InTech

MAR-APR 2018

Issue link: http://intechdigitalxp.isa.org/i/959322

Contents of this Issue

Navigation

Page 43 of 57

44 INTECH MARCH/APRIL 2018 WWW.ISA.ORG AUTOMATION BASICS If the deviations are too slow in na- ture to be considered noise, a filter should not be applied. Because the PV deviates relatively slowly from its SP, the controller may seem to respond to it too slowly, allowing large deviations before gaining control over the process. This is likely a result of sluggish tun- ing, which calls for the controller to be tuned again for a faster response if pos- sible (as explained later). Note that deadband in the control valve may create the appearance of the controller responding sluggishly. It takes a while for the controller to traverse the deadband before the control valve actu- al ly starts responding and affecting the process. In this case, controller tuning is not the right solution for the problem, since deadband is a mechanical issue with the control valve. The presence of deadband in a control valve can be deter- mined by putting the controller in manu- al mode and making two CO changes in one direction, and a third change in the opposite direction. Make the second and final steps equal in size and wait for the PV to settle out after every step. If, after the final step, the PV does not return to the same value as after the first step, dead- band ex ists in the control valve (figure 5). Also realize that every control loop has a limit on its speed of response. If this limit is exceeded by tuning the controller for too fast of a response, the control loop will become unstable and oscillate. Best practices for controller tuning recommends leaving a healthy margin away from the stability limit. If a controller has been tuned prop- erly, and deviations of PV from the SP are still excessive, it may be the result of the disturbances being too severe to be handled solely by feedback control. Depending on the cause of the distur- bance, adding cascade, feedforward, or ratio control to the control design could make a significant improvement. n ABOUT THE AUTHOR Jacques Smuts, PhD, PE (jsmuts@opticon- trols.com), is the founder and principal consultant of OptiControls, Inc. He has more than 25 years of experience in pro- cess control and is the author of the book Process Control for Practitioners. is put in manual, both control loops cease to oscillate. In this case, the most impor - tant control loop needs to be tuned for a fast response, and the other loop tuned to respond three to five times slower. The lambda tuning method lends itself well to tuning the second loop for a slow response. Random deviations Now on to the case in which the PV devi- ates from its SP in a random (nonoscilla- tory) fashion. If the deviations occur very rapidly, much faster than the control loop can respond, the deviations can be considered to be noise. The best solution for PV noise is to apply a small filter, such as a first-order lag filter, to the signal. Note that a filter changes the dynamics of the control loop, and consequently the controller has to be tuned again. closely resembles a triangular wave, while the PV more closely resembles a square wave (figure 4). Control valve st - iction is a common cause of control loop oscillations, and being a valve problem, it needs to be addressed through control valve maintenance or positioner tuning. When incorrect tuning settings cause a control loop to oscillate, the PID con - troller needs to be tuned again, prefer- ably by using a scientific tuning method, not trial and error. Another scenario to consider is when two or more closely coupled control loops cause each other to oscillate. For example, if a pressure-reducing control loop is close - ly followed in the process by a flow control loop, the two control loops can appear to be fighting each other, causing both loops to oscillate. If either one of the controllers Figure 5. Control valve deadband test Figure 4. Control loop oscillation caused by stiction

Articles in this issue

Links on this page

Archives of this issue

view archives of InTech - MAR-APR 2018