The objectives of process control are generally either to maintain a process at a desired, constant operating condition (temperature, pressure, composition, etc.) in the face of disturbances or, less typically in conventional process applications, to force it to follow a desired trajectory with time.
All process control configurations, whether manual, automatic, or computer-based, have three essential elements:
- a measurement (often several);
- a control strategy (embedded in a controller);
- a final element for implementing the control action (a valve, heater or other variable input).
For example, in classical feedback control , one can correct for the effects of a disturbance by:
- measuring some property of the system to be controlled (its temperature, for example):
- comparing it with its desired value (set point) and calculating what changes, if any, are required as inputs to the process;
- making an appropriate adjustment to that/those inputs (a variable heater, for example).
Similarly, in feedforward control , one can anticipate and compensate for the effects of a major disturbance to the process by first measuring the disturbance; this would then be followed by the other two steps described above. Both feedback and feedforward control can be used on the same process as illustrated diagrammatically in [Figure 1]
In a conventional (analogue) control environment, the measurements, controller calculations and signals to the final elements are typically carried out by electrical signals, generally in the range of 4-20 mA; in more modern computer control systems, any analogue measurements are converted to digital form for processing and the resulting signal to the final element is reconverted to analogue form where necessary.
A wide range of on-line measuring devices (sensors) exists in the process industries, the most common being those for flow rate, pressure, liquid level, temperature, pH, and other selective measures of chemical composition.
Classical PID (proportional-integral-derivative) controllers still dominate feedback strategy in industrial applications. Here, the signal to the final control element, p, is related to its nominal value, ps, and to the “error” signal, e, by the equation
where Kc, τ1, and τD are the proportional, integral, and derivative constants, respectively.
The control action signal calculated by the controller is sent to the final control element, a device which implements the change of a suitable input to the system. This variable input is typically a flow rate (of fuel, air, coolant, reactant, etc.). Hence, an automatic control valve, either electrically or pneumatically operated, is usually appropriate.
Typical problems which lead to the poor performance of control systems include: inadequate/inaccurate measurements; long time delays in either the process or the measurements; varying or nonlinear nature of the system; the interaction of several quasi-independent control loops.
The major development over the past two decades has been the replacement of analogue controllers by digitally calculated control signals, even for conventional PID controllers. In addition, there has been considerable effort to develop more sophisticated model-based predictive controllers.
In a conventional controller, the controller parameters are often adjusted (“tuned”) on-line, although prior knowledge of the system to be controlled will provide some guidance in their choice. Model-based control, on the other hand, makes use of a model of the process (either fundamentally-based or empirical) to decide on the appropriate control strategy to be implemented to achieve the desired objective. In principle, these can lead to improved performance, but, in practice, their performance may be strongly dependent on the accuracy of the model used.