Global Structure Evolution and Local Parameter Learning for Control System Model Reductions
This chapter develops a Boltzmann learning refined evolution method to perform model reduction for systems and control engineering applications. The evolutionary technique offers the global search power from the ‘generational’ Darwinism combined with ‘biological’ Lamarckism. The evolution is further enhanced by interactive fine-learning realised by Boltzmann selection in a simulated annealing manner. This hybrid evolution program overcomes the well-known problems of chromosome stagnation and weak local exploration of a pure evolutionary algorithm. The use of one-integer-one-parameter coding scheme reduces chromosome length and improves efficiency dramatically. Enabled by a control gene as a structural switch, this indirectly guided reduction method is capable of simultaneously recommending both an optimal order number and corresponding parameters. Such order flexibility and the insight it provides into the system behaviour cannot be offered by existing conventional model reduction techniques. The evolutionary approach is uniformly applicable to both continuous and discrete time systems in both the time and the frequency domains. Three examples involving process and aircraft model reductions verify that the approach not only offers higher quality and tractability than conventional methods, but also requires no a priori starting points.
KeywordsModel Reduction Order Number Pseudo Random Binary Sequence Model Reduction Method Model Reduction Problem
Unable to display preview. Download preview PDF.
- 2.Anderson, J. H., 1967. ‘Geometric approach to the reduction of dynamical systems’, IEE Proc. Control Theory Appl., Pt. D, 114(7), 1014–1018.Google Scholar
- 8.Kwong, S., Ng, A. C. L., and Man K. F., 1995. ‘Improving local search in genetic algorithms for numerical global optimization using modified GRID-point search technique’, Proc 1 st IEE/IEEE Int. Conf. on Genetic Algorithms in Eng. Syst.: Innovations and Appl., Sheffield, UK, 419–423.Google Scholar
- 10.Li, Y., 1996. Neural and Evolutionary Computing IV, Fourth Year Course Notes (1JYX), Department of Electronics and Electrical Engineering, University of Glasgow.Google Scholar
- 12.Li, Y., and Ng, K. C., 1996. ‘A uniform approach to model-based fuzzy control system design and structural optimisation’, in Genetic Algorithms and Soft Computing, F. Herrera and J. L. Verdegay (Eds.), Physica- Verlag Series on Studies in Fuzziness, Vol. 8, 129–151.Google Scholar
- 15.Ng, K. C., 1995. Switching control systems and their design automation via genetic algorithms, Ph.D. Thesis, Department of Electronics and Electrical Engineering, University of Glasgow.Google Scholar
- 16.Renders, J. M., and Bersini, H., 1994. ‘Hybridizing genetic algorithms with hillclimbing methods for global optimisation: two possible ways’, Proc 1 st IEEE Int. Conf. Evolutionary Computation, First IEEE World Cong. Computational Intelligence, Orlando, 1, 312–317.Google Scholar
- 20.Xue, D., and Atherton D. P., 1993. ‘An optimal reduction algorithm for discretetime systems’, Proc. IFAC 12th Triennial World Congress, Sydney, Australia, 821–824.Google Scholar