The approach to design and common pitfalls

Designing a control system for a complete plant can involve the specification of several hundred control loops along with trips, alarms, overrides and all sorts of extra bits and pieces. It is a huge task and requires considerable experience to do well. The notes I've but together here won't make you a control system designer, but should give you some hints on how to tackle the control aspects of your design project.

Putting together a control specification

The first stage in design is to put together a specification which details exactly what you are trying to achieve. One method of doing this is to follow the sequence below:

1. Produce a specification for the overall process. Define the required outputs (usually production rates - nominal value and variability, quality, etc). Specify constraints - most usually environmental constraints on discharges. Identify and quantify likely disturbances which will affect the whole process (e.g. know variations in raw materials, cooling water temperatures, etc).
2. Take the flowsheet of the process and start to divide it into a series of function units. This decomposition is quite similar to breaking a process into unit operations for process design, but functional units for control can often include several process units (e.g. it makes sense to include the reboiler, condenser and reflux accumulator in a distillation column functional unit)
3. Now take the process control specification and try to decompose it into a number of more detailed specifications for each function unit. This will involve translating global objectives, like production rate and quality, into unit objectives involving individual stream flowrates, temperatures, pressures, etc.  It is important to remember that in control its the allowable variation in these quantities that should be included in the specification. Do the same with the constraints and disturbances. Extra constraints (e.g. pressure limits on equipment) and disturbances (e.g. likely cooling effect of rain falling on a distillation column) will also appear as this process proceeds. Identify what manipulations are available and their ranges.
4. Repeat steps 2 and 3 several times. Often the first go at dividing a flowsheet into functional units is the best, but it's always worth looking to see what happens if a piece of the process (e.g. a heat exchanger) is moved from one unit to an adjacent one.
5. Now chose appropriate loop pairings to link the manipulated and controlled variables
6. Finally add any 'bells and whistles' (e.g. overrides, cascades, trips, etc) that may be required.

Common pitfalls in design projects

Here are some of the most common mistakes which are made in the control parts of the design project:

• Putting two control valves on the one line. This is a nice way to deal with limited degrees of freedom, but unfortunately it won't work! At least one valve will quickly go fully closed or fully open and control will be lost on at least one of the variables. The following won't work either:
  • Taking a spur around a control valve, inserting a control valve in this, and then rejoining the spur to the main line. The flow that's being manipulated is the sum of the flow in the main line and the spur - adding the spur doesn't add a degree of freedom (NB: adding a spur can be appropriate when a high degree of rangeability is required on the flow, the spur can contain a small valve for low flow adjustments. Configuring such a system can be a bit awkward!)
  • Putting a control valve on the entry and exit lines from a heat exchanger. Although a heat exchanger is a piece of equipment, it doesn't (on the tube-side at least) accumulate liquid - its just a big bit of folded pipework. Attempting to manipulate the inlet and outlet flows is just putting two valves on the same bit of pipe. You can manipulate the inlet and outlet flows from a tank independently because liquid can accumulate inside the tank (and act as a reservoir).
• Not controlling inventories. You must control the liquid level (the liquid inventory) in every vessel that has a liquid level. If you don't the vessel will drain completely or overflow, depending on process conditions. In enclosed systems you also must control the pressure in vessels (the gas inventory). Inventories must be controlled, even at the expense of losing degrees of freedom for controlling production and quality variables. If you really need to control these and don't have a degree of freedom available then you need to redesign the process, or take the degree of freedom from another functional unit (i.e. redo the decomposition of the flowsheet)
• Putting control valves on the suction lines of pumps. If you do this the pump will cavitate when the valve closes (the suction pressure will drop below the cavitation pressure) and will fall to bits. Note, however, that the normal way to control centrifugal compressors is to put a valve in the suction line - compressors don't cavitate.
• Putting a valve in the discharge line of a positive displacement pump. These pumps always deliver the same flowrate, and if you try to throttle them the line pressure will quickly increase until something bursts! The correct way to deal with these is to put the control valve on a 'kick-back' line (a line which connects the discharge to the inlet). As the control valve opens more of the flow is directed back into the pump inlet and less goes down the process line. Dosing pumps often directly accept control signals which alter the stroke length or frequency and avoid the need for a kick-back line and control valve.
• Trying to control things that can't be easily measured. Concentrations and other quality variables can be difficult to measure and in some cases the only measurement may be from a laboratory analysis available only once every few hours. In such cases it is necessary to control the process on other variables which are related to the concentration (e.g. temperature and flows) and adjust the setpoints on these as laboratory analysis becomes available. Before suggesting something as a controlled variable always think about how you are going to measure it.