Process Control System Design for a Distillation Unit

The aim of a typical control system is to force a given set of process variables to act in some desired and prescribed way by either fulfilling some requirements of the time or frequency domain or achieving the best performances as expressed by an optimization index. The scope of the control tasks varies widely. The key goal may be to keep the process running around the nominal conditions. In other cases, the control objective will be to transfer the plant from one operating point to another or to track a given reference signal. In some other cases, the interest may lie in obtaining the best features of the plant, achieving, for example, the maximum production, minimum energy consumption or pollution or minimum time in performing a given task.

Generally speaking, the following control objectives can be targeted:

• Regulation (disturbance rejection).
• Reference tracking.
• Generation of sequential procedures (for start-up or shut-down).
• Adaptation (changing some tunable parameters).
• Fault detection (to prevent process damage or provide reconfiguration).
• Supervision (changing the operating conditions, structure or components).
• Coordination (providing the set-points).
• Learning (extracting some knowledge from the experience).

All the above mentioned activities result in very distinct control approaches and techniques. From logical and discrete time controllers to complex intelligent control systems where AI techniques provide the framework for emulating human behaviour, the various available tools for control systems design are complementary and in the integral control of a plant, some being used in a cooperative way.

Process Control System Design for a Distillation Unit: Key Considerations

Let us consider a typical process unit for refining a chemical product as illustrated in the figure below.

Process, Instrumentation and Variables

In reference to the above figure, first, there is a mixing of two raw materials (reactives) to feed a distillation column where two final products are obtained, the head and the bottom components. In order to run the unit, we must control the different flows of material, provide adequate temperature to the inlet flows and keep the desired operating conditions in the column by adjusting its temperature, pressure and composition. Some other complementary activities are necessary such as agitating the content of the mix tank or keeping the appropriate levels (of course accomplished with level controls) in all vessels, including those of auxiliary or intermediate buffers.

The ultimate control objective is to achieve the best distilled products (maximum purity, less variance in concentration, and so forth) under the best conditions (maximum yield, minimum energy consumption, etc.), taking into consideration the cost and pollution constraints. Before we start getting products out of the system, we must startup all the equipment devices, establish a regular flow of reactives, reach the nominal operating conditions and then keep the unit stable under production. Furthermore, care should be taken about faults in any part of the unit i.e. valves, agitator, existence of raw materials, heating systems, etc.

Independently of its design, carried out taking into account requirements or not, control design assumes that the equipment modules are given and are already interconnected according to the guidelines of the process experts. At times, analysis of expected performance with a particular control system may advise changes in the process or instrumentation (sensors and actuators); for example, from the control viewpoint, it could be better to feed the mixed material or the reflux of the head product at a different column plate, however at this moment it is fixed.

To control the process, some manipulated variables should be available, allowing the introduction of control actions in the process to force it towards evolving in the desired way. In an automatically controlled plant, these manipulated variables will act on the process through the corresponding actuators. To get information about the process, some internal variables should be measured, being considered as output variables. Again, more than one output variable will be considered. The control target could be these variables themselves or some other directly related to them: to keep them constant in a regulatory system, to track some references in servo systems or to perform in some prescribed way with temporal, harmonic or stochastic properties.

In the distillation unit (Figure 1 above), there are several input variables (14 valves, 2 pumps and an agitator). All of them can be used to drive the unit, being considered as manipulated variables, however, most of them will be locally controlled or manually fixed and will not intervene in the control strategy. The temperatures, flows, levels or concentrations at different points inside the unit can provide information about the behaviour of the plant, but not all of them will be measured, much less controlled. The set of measurement devices, as well as the instrumentation needed to condition the measurements, constitute the data acquisition system, which itself can be quite sophisticated involving transducers, communication lines and converters. These devices are fundamental in achieving proper control.

The input variables or signals acting on the process but not being manipulated to achieve the control goals should be considered as disturbances. They are usually determined as a result of other processes or, in the simplest case, they are assumed to be constant. These disturbances can be predictable (deterministic) or not. For example, in a rolling mill process, the arrival of a new block affecting the roll’s speed is an event that can be predicted but not prevented. Also, the disturbances can be measurable or not. Still some characteristics of the disturbance may be known in advance if it belongs to a class of signals. For instance, in the distillation column, we can get the information about the raw materials’ concentration, but it is fixed somewhere else and cannot be considered as a manipulated variable. Ambient temperature is also a partially predictable disturbance.

It is obvious that unpredictable, unknown and unmeasurable disturbances are the worst ones to be counteracted by the control actions.

Generally process variables can be classified as:

External Variables or Inputs: They are determined by other processes or the environment, acting on the process and considered as manipulated or disturbances. Manipulated/control variables (if used to influence the dynamics of the process), actuators will amplify the control commands to appropriate power levels to modify the plant’s behaviour. Disturbances (if they are uncontrollable outputs of other subsystems).

Internal Variables: They are dependent on the process inputs, system structure and parameters. They can be classified as outputs or measured variables, controlled variables, state variables. Outputs/measured variables, if they are sensed and provide information about the process evolution. Controlled variables, if the control goals are based on them. They can be outputs or not, depending on the sensor’s availability and placement. State variables, are a minimum set of internal variables allowing the computation of any other internal variable if the inputs are known.

Most of the distillation unit variables are modelled using continuous time (CT) signals, however, for the sake of control, could be treated digitally or logically. For example, the agitator speed could be represented by two options: ON and OFF. Some variables could be treated as continuous time signals if they lie inside a prescribed range of values, being considered as saturated or null if they are out of range. That is to say, the same physical variables can be represented by different signals depending on the purpose of their treatment.

In some cases, the value taken by a variable as a function of time is not relevant and we may be interested in some periodic properties, such as the frequency components, in magnitude or phase. Harmonic analysis (Fourier transform) is thus suitable. In some other cases, only the stochastic properties of the variables are of interest. For instance, consider the concentration of a distillation product. More than the prompt value of the concentration at a time instance, the interest of the user is in the average concentration in a reasonable interval, as well as the possible deviations.

Related articles:

• Basic Steps to Consider in Designing a Control System
• Steam-heated Reboiler Control System
• Principles of Control Systems
• Integral (Reset) Control Systems
• Proportional plus Derivative (PD) Control System
• Furnace Control System for a Rapidly Changing Demand