Instrumentation Reference Guide--Measurement Methods and Control Strategies

1 Introduction

Measurement methods for automation are somewhat different from those designed for use in laboratories and test centers.

Specific to automation, measurement methods that work well in a laboratory might not work at all in a process or factory-floor environment. it might not be possible to know all the variables acting on a measurement to determine the degree of error (uncertainty) of that measurement. For example, a flowmeter may be calibrated at the factory with an accuracy of ±0.05% of actual flow. Yet in the field, once installed, an automation professional might be lucky to be able to calibrate the flowmeter to ±10% of actual flow because of the conditions of installation. Because control strategies are often continuous, it’s often impossible to remove the sensor from the line for calibration.

2 Measurement and Field Calibration Methodology

In many cases, then, field calibration methods are expedients designed to determine not the absolute accuracy of the measurement but the repeatability of the measurement.

Especially in process applications, repeatability is far more critical to the control scheme than absolute accuracy. it’s often not possible to do more than one or two calibration runs in situ in a process application. it often means that the calibration and statistical repeatability of the transmitter is what is checked in the field, rather than the accuracy of the entire sensor element.

3 Process control strategies

Basic process control strategies include on/off control; deadband control; proportional, integral, derivative (PID) control; and its derivatives.

On/off control is simple and effective but may not be able to respond to rapid changes in the measured variable (known as PV, or process variable). The next iteration is a type of on/off control called deadband, or hysteresis control.

In this method, either the "on" or the "off" action is delayed until a prescribed limit set point is reached, either ascending or descending. often multiple limit set points are defined, such as a level application with "high" level, "high-high" level, and "high-overflow" level set points. Each of the set points is defined as requiring a specificaction.

Feedback control is used with a desired set point from which deviation is not desired. When the measured variable deviates from the set point, the controller output drives the measured variable back toward the set point. Most of the feedback control algorithms in use are some form of PID algorithm, of which there are three basic types: the standard, or ideal, form, sometimes called the ISA form; the inter active form, which was the predominant form for analog controllers; and the parallel form, which is rarely found in industrial process control.

In the PID algorithm, the proportional term provides most of the control while the integral function and the derivative function provide additional correction. in practice, the proportional and integral terms do most of the control; the derivative term is often set to 0.

PID loops contain one measured variable, one controller, and one final control element. This is the basic "control loop" in automation. PID loops need to be "tuned"; there are several tuning algorithms, such as Ziegler-Nichols and others, that allow the loop to be tuned. Many vendors today provide automatic loop-tuning products in their control software offerings.

PID feedback controllers work well when there are few process disturbances. When the process is upset or is regularly discontinuous, it’s necessary to look at other types of controllers. Some of these include ratio, feed forward, and cascade control. In ratio control, which is most often found in blending of two process streams, the basic process stream provides the pacing for the process while the flow rates for the other streams are modulated to make sure that they are in a specific ratio to the basic process stream. Feed forward control, or open loop control, uses the rate of fall-off from the set point (a disturbance in the process) to manipulate the controlled variable. An example is the use of a flowmeter to control the injection of a chemical additive downstream of the flowmeter. There must be some model of the process so that the effect of the flow change can be used to induce the correct effect on the process downstream.

Combining feed forward and feedback control in one integrated control loop is called cascade control. In this scheme, the major correction is done by feed forward control, and the minor correction (sometimes called trim) is done by the feedback loop. An example is the use of flow to control the feed of a chemical additive while using an analyzer downstream of the addition point to modulate the set point of the flow controller.

4 Advanced control strategies

Since the 1960s, advances in modeling the behavior of processes have permitted a wholly new class of control strategies, called advanced process control, or APC. These control strategies are almost always layered over the basic PID algorithm and the standard control loop. These APC strategies include fuzzy logic, adaptive control, and model predictive control.

Conceived in 1964 by University of California at Berkeley scientist Lotfi Zadeh, fuzzy logic is based on the concept of fuzzy sets, where membership in the set is based on probabilities or degrees of truth rather than "yes" or "no."1. Because:

1. Britannica Concise Encyclopedia, quoted in www.answers.com multiple fuzzy logic sets appear to be able to learn, they are often regarded as a crude form of artificial intelligence. In process automation, only four rules are required for a fuzzy logic controller:

2 Rule 1: If the error is negative and the change in error is negative, the change in output is positive.

Rule 2: If the error is negative and the change in error is positive, the change in output is zero.

Rule 3: If the error is positive and the change in error is negative, the change in output is zero.

Rule 4: If the error is positive and the change in error is positive, the change in output is negative.

Adaptive control is somewhat loosely defined as any algorithm in which the controller's tuning has been altered. Another term for adaptive controllers is self-tuning controllers.

Model predictive control uses historicized incremental models of the process to be controlled where the change in a variable can be predicted. When the MPC controller is initialized, the model parameters are set to match the actual performance of the plant. An expert notes: "MPC sees future trajectory based on past moves of manipulated variables and present changes in disturbance variables as inputs to a linear model. It provides an integral-only type of control." These advanced control strategies can often improve loop performance but, beyond that, they are also useful in optimizing performance of whole groups of loops, entire processes, and even entire plants themselves.

PREV. | NEXT