Introduction
Inferential control is a control scheme in which the controlled variable (primary process variable) is not measured directly but estimated from a secondary measurement.
The secondary measurement variable is termed as an inferential variable. The estimated value of the controlled variable (primary process variable) is compared against the set point and the control action is taken. In this control scheme, the process disturbances are also not measured.
When to employ Inferential Control
1. Direct measurement of the controlled variable (primary process variable) is not possible because
- An online sensor is not available,
- The available online sensor is very expensive to be installed and maintained,
- The sensor needs human management in providing accurate and reliable measurements which makes it not suitable for real-time measurements,
- The sensor has unfavorable dynamics like long dead time, long processing time.
3. A measured inferential variable is available.
Inferential Variable
The inferential variable is a variable that is closely related to the unmeasured controlled variable. The value of the controlled variable is concluded (estimated) from the measurement of the inferential variable.
Example: The process variable that is most commonly inferred from the secondary measurement is composition. There is a lack of reliable, rapid and economic measuring devices for the measurement of the composition. Also, temperature provides a valuable inference of composition. Temperature is the most common inferential variable used to infer the unmeasured composition.
Conditions for the Inferential Variable
1. The inferential variable must have a good relationship with the controlled variable for changes in the manipulated variable.
2. The relationship between the inferential variable and the controlled variable must not change with the process disturbances over its operating range.
Structure of Inferential Control
The process (shown in figure 1) that needs inferential control to be employed has one unmeasured controlled variable, y and one secondary measurement of an inferential variable, z. The manipulated variable, m, and the disturbance d affect the process and influence the outputs. Gp1,Gp2, and Gd1,Gd2 are the transfer functions of process and disturbances respectively. It is assumed to be that the process model is known and so the transfer functions are known. The input-output relationships of the process are derived as follows.
Structure of Inferential Control
 |
| Figure 1: Process that needs Inferential Control |
y = Gp1m + Gd1d ....(1)
z = Gp2m + Gd2d ....(2)
Solve equation (2) for d to find the estimate of the unmeasured disturbance.
Substitute equation (3) into equation (1) to find the estimator of the unmeasured controlled variable from the known manipulated variable and inferential variable.
 |
| Figure 2: Structure of Inferential Control System |
Limitations of Inferential Control
1. The measurement of the inferential variable is not as accurate as an online sensor of the controlled variable. Also, the approximate relationship used for the inferential variable has a limited range, beyond which the inferential variable might not be satisfactory.
2. The success of inferential control depends on the knowledge of the process. The performance of inferential control depends on a good estimator which is constructed using process transfer functions. Usually, the process transfer functions are known approximately which affects the estimator and in turn the quality of inferential control.
Further Reading: Applications of Inferential Control
Also, read other control schemes
Further Reading: Applications of Inferential Control
Also, read other control schemes
No comments:
Post a Comment
Share your reading experience here