A comparison of binary coding and real-coded genetic algorithm applied to a three-phase catalytic reactor

Systematic methods and tools for managing the complexity

Process Simulation and Optimization (T4-9P)

MSc Mylene Rezende
State University of Campinas
Department of Chemical Engineering
Cidade Universitária Zeferino Vaz, CP 6066, CEP 13087-970 Campinas-SP, Brazil
Brazil

Dr Aline Costa
State University of Campinas
Department of Chemical Engineering
Cidade Universitária Zeferino Vaz, CP 6066, CEP 13087-970 Campinas-SP, Brazil
Brazil

Prof Rubens Maciel Filho
State University of Campinas (UNICAMP)
Department of Chemical Processes, School of Chemical Engineering
Zeferino Vaz, University City
Barrão Geraldo - Campinas - São Paulo - Brazil

Laboratory of Optimization, Design and Advanced Control.
Department of Chemical Processes. School of Chemical Engineering.
State University of Campinas (UNICAMP)


PO BOX 6066
ZIP 13081-970
Brazil

Keywords: Optimization, Genetic Algorithms, binary coding, real-coded, three-phase reactor

Genetic Algorithms have been often applied to solve many optimization problems. Such algorithms are based on Natural Genetic and Natural Selection mechanism and some fundamental ideas are borrowed from Genetic in order to artificially construct an optimization procedure. The Genetic Algorithms start with a population of possible solutions, which suffers evolution during the generations. Each solution is coded as a set of binary or real strings (chromosome), each string representing a variable in the solution. The evolution occurs when some genetic operators as reproduction, crossover and mutation are applied. The survival of the fitness is achieved by the assignment of a fitness function. The classical form of Genetic Algorithm used to solve an optimization problem is a binary coding which works with binary strings. There is another form that uses the real values directly that is the real-coded Genetic Algorithm. The real-coded Genetic Algorithm has been used preferentially instead of the binary coding because the real-coded Genetic Algorithm does not depend on the arbitrary precision in optimization variables and because during the procedures of binary coding for a real number may occur loss precision depending on the bits number used. This paper aims to compare the binary and real-coded Genetic Algorithm in respect to time consuming and the best value of the objective function, in order to point the suitable approach for a Real-Time optimization. That comparison is applied to a three-phase catalytic reactor that produces 2-methyl-cyclohexanol and the problem is postulated as a single objective and constrained optimization problem.

Presented Wednesday 19, 13:30 to 15:00, in session Process Simulation and Optimization (T4-9P).