dc.description.abstract | Advances in digital computing over the last years have resulted in new and powerful tools for
obtaining process models. An example of such a tool is the dsr toolbox, which gives a state
space model based on measured input/output data. Also, new control strategies based on
these models have developed, usually involving optimization techniques. Despite this, the
classical PID controller still has advantages and remain the most used control technique.
The goal of the thesis was to compare different methods for tuning PID controllers. The
advantages and disadvantages of the different methods should be explained and suggestions
of how the methods could be used with state space models should be discussed. The Matlab
pidtune function and the delta tuning rules should be explained and evaluated in relation
to state space models.
The tuning methods Ziegler Nichols, SIMC, Cohen-Coon, and optimization tuning in addition to δ-tuning and pidtune, was chosen to examine in detail. To obtain model parameters
for controller tuning from state space models, a graphical method, an optimization method
and the Matlab function procest was used. Pidtune, mftune, megatuner, and optimization
based tuning is used directly with SSM and was also tested. For method comparison, both
commonly known process models and random models were used.
The methods which can be used directly on state space models give the best results in terms
of successful tuning attempts. For many higher order SSM, process describing variables such
as K, θ , T, R, and L can be found successfully by graphical estimation or optimization.
These variables are then used for PID controller tuning. The graphical method is the fastest
and gives the highest success-rate, while optimization estimation results in higher closed-loop
performance | en_US |