Project Title: Multi-objective Optimisation of a PID Controller
Supervisor:Tim Hendtlass (overall) & Dan Angus (day-to-day)
Suitable year level:3rd-5th year
Project Description
In control system design, proportional-integral-derivative (PID)
controllers are an industry standard due to their simple structure, wide
range of operating conditions and robust performance. While many
heuristics are available to tune these controllers for specific
applications (e.g., Ziegler-Nichols method), the Engineer is often left
wondering whether the configuration is in fact the best for the purpose.
The application of Evolutionary Computation algorithms to the tuning of
PID controllers has been met with success although these approaches tend
to group quality measures such as Rise Time, Overshoot, Settling Time,
Steady-State Error and Stability into a single measure of quality.
Multiple Objective Optimisation (MOO) is concerned with finding multiple
`trade-off' solutions in order to optimise many (in most cases conflicting
or orthogonal) objectives. For all MOO problems there is a set of optimal
trade-off solutions which are referred to as the Pareto set, after the
economist Vilfredo Pareto. To be classified as Pareto optimal a solution
must not be worse then any other valid solution in all objectives. A
Pareto optimal solution cannot increase its quality in any objective
without simultaneously decreasing its quality in another objective.
This project involves the modelling of a basic closed-loop control system
as a MOO problem where each quality measure is treated independently. Once
correctly modelled the student will be free to design their own algorithm
or apply any of the already existing state-of-the-art MOO algorithms with
the intent to find Pareto optimal solutions for a variety of plant
configurations such as an electric motor speed controller.
Expectations/Assessment
A technical report describing the various parameters of the hardware/software and the research that underpins the final product.
Pre-requisite Knowledge
The work can be completed using any software tools familiar to the student
(e.g. MATLAB[tm], Java[tm]). It is believed that if completed to a
suitable standard this work would be eligible for publication.
Further details:dangus @ ict.swin.edu.au
Back |