Control Systems Design Routes

In terms of improving the performance of control systems, there are four stages that can be followed (generally in sequency) which addresses all of these levels:

  • stage 0 - system risk reduction (safety). For level 0 and 1 control systems.
  • stage 1 - monitoring and reporting (including Statistical Process Control). This primarily addresses level 1 and 2 control systems
  • stage 2 - benchmarking against internal or compeititor systems. This primarily addresses level 1 and 2 , but can also be applied to level 3 control systems.
  • stage 3 - overall system optimisation, through data mining techniques. Can look at level 1-3 control systems, but level 1 systems should have been optimised through earlier stages.

Studies to look at control system improvements can involve a number of tools and routes, all of which offer benefits to particular applications.

  • simulation studies using computer aided control system design tools (CACSD) - these can be used to look at a very wide range of control issues such as new techniques and what-if scenarios
  • commissioning and improvement tools - primarily for re-tuning existing loops
  • hardware-in-the-loop - development systems where actual control hardware systems are prototyped with a plant simulation
  • rapid prototyping systems such as DSPACE for Matlab, AC100 for Matrixx, UNAC etc.

Links on Design procedures

  • Ken Carter (President, Control Technology Corp) is writting an article on control system development practices for embbedded controllers. This will be linked in here when ready.

Modelling and Simulation

Most control system designs rely on modelling of the system to be controlled. This allows simulation studies to be carried out to determine the best control strategies to implement and also the best system parameters for good control (if these can be changed). Simulation studies also allow complex what-if scenarios to be looked at which may be difficult to do on the real system. However, it must always be borne in mind that any simulation results obtained are only as good as the model of the process. This does not mean that every effort should be made to get the model as realistic as possible, just that it should be sufficiently representative.

Here are a few links on modelling in general:

The most common way to construct a model is using a combination of transfer functions and nonlinear elements to represent the mathematical equations that decsribe the system. Here are a few links on block diagram modelling:

  • Here's a simple diagram that shows the fundamental block diagram manipulations for combining blocks, moving pick-off points and summing nodes and eliminating feedback loops.

Bond graphs are a method of modelling physical dynamic systems that do not rely on the complex mathematical descriptions often used for developing model from physical principles and equations. The following links relat to bond graphs: