Particle swarm optimisation algorithms and their application to controller design for flexible structure systems

Particle swarm optimisation (PSO) is one of the relatively new optimisation techniques, which has become increasingly popular in tuning and designing controllers for different applications. A major problem is that simple PSO have a tendency to converge to local optima, mainly, due to lack of diversity in the particles as the algorithm proceeds and improper selection of other parameters. Maintaining diversity within a population is challenging for PSO, especially for dynamic problems. In order to increase diversity in the search space and to improve convergence, a new variant of PSO is proposed. The increased interest from industry and real-world applications has led to several modifications in the conventional algorithms so as to deal with multiple conflicting objectives and constraints. A modified multi-objective PSO (MOPSO) proposal is made which will allow the algorithm to deal with multi-objective optimisation problems. The main challenge, in designing a MOPSO algorithm, is to select local and global best for each particle so as to obtain a wide range of solutions that trade-off among the conflicting objectives. In the proposed algorithm, a new technique is introduced that combines external archive and non-dominated fronts of the current population in order to select the global best for each particle. The effectiveness of the proposed algorithm is assessed with two examples in controller design for vibration control of flexible structure systems and satisfactory results have been obtained.