Evolving Population-Based Search Algorithms through Thermodynamic Operation: Dynamic System Design and Integration

Abstract

During the last three decades there has been a growing interest in algorithms which rely on analogies to natural processes. The guided random search techniques, genetic algorithms (GAs) and simulated annealing (SA), are very promising strategies, and both techniques are analogs from physical and biological systems. These two algorithms are stochastic relaxation search methods especially, suitable for applications to a wide variety of complex optimization problems. Each produces a sequence of candidate solutions to the underlying problems, and the purpose of both algorithms is to generate sequences biased toward solutions which optimize the objective function.Limitations, however, occur in performance because optimization may take a large number of iterations, and final parameter values may be found that are not at the global extremum points. In this thesis, a population-based search algorithm that combines approaches from GAs and SA is proposed. The combined approach, called GASA, maintains a population of individuals over a period of generations. In the GASA technique, simulated annealing is used in choices regarding a subset of individuals that undergo crossover and mutation. This thesis shows that the GASA technique has superior performance when compared to a genetic algorithm for the nonlinear function optimization problem.Temperature plays an important role in the GASA algorithm, The temperature change results in the moving of populations. When arriving at equilibrium, the individuals in the same population also have an assumed unique critical temperature. This phenomenon follows the rules of general thermodynamic laws. Schema theory and simulated phase transition are used to explore the search mode of GASA. Energy changes affect structural change by mutation or crossover. GASA is utilized for parameter optimization on dynamic system design and integration. The GASA technique can be used for different tasks like control, detection, and computation. For fed-batch bioprocesses, an optimized input function (substance feed rate) with GASA can increase the quantity of product. GASA is applied to pH control in combination with the classical PID control. A convergence analysis of GASA explores the characteristics, of this evolutionary controller.