Computational Intelligence And Its Application To Engineering

Computational intelligence (Cl) is a computing methodology from nature, which simulates and researches the intelligent behavior from the lowest level of the creature. CI develops the traditional style of computation and provides a new approach for solving complex problems.In order to enhance the applying efficiency of CL, the cause of premature convergence in binary-coded genetic algorithms (GAs) is analyzed in this dissertation. The drawback of conventional mutation operator in preventing premature convergence is subsequently pointed out. Whereafter, a genetic algorithm, which can be implemented via general logic gate circuit, is proposed. Considering the influence of parameters selection on the solving efficiency of GAs, the influence of parameters of GAs on its performance is analyzed by means of schema theorem presented by Holland, which provides a theoretical warrant for the selection of parameters. In research of parallel GM, a parallel GAs based on the marriage principle (PGAMP) is put forward inspired by the marriage rule of human being. The experiments show that PGAMP possesses a high search efficiency. Learning algorithm is the bottleneck in employing artificial neural networks (ANNs). This dissertation analyzes the problems of learning rate adaptation in the learning algorithms of multi-layer feedforward ANNs and then proposes a scheme of adapting learning rate. Experimental test demonstrates that the scheme can universally accelerate the search speed.In the engineering application of CI, Two methods of evolutionary computing and neural computing the Fourier factors are proposed which redound to the application of Fourier transformation to the engineering. In the area of automatic control, a novel conception of online designing controller is brought forward in this dissertation. An online evolutionary design scheme of ITAE-optimal-control-law based neural controller is presented after the combination of evolutionary computation (EC) and ANNs. Simulation results display the feasibility of the scheme. In the application of fuzzy logic to the engineering, the dissertation presents a new approach to remove the steady-state error of fuzzy logic controller (PLC) without using any integrator. The prominent character of the approach is that the FLC can on-line search for the steady-state control value needed for the zero-error tracking, and optimize the fuzzy control table dynamically. The feasibility of the approach is tested in the control of CSTR.