Open AccessBinary Coded Genetic Algorithm (BCGA) with Multi-Point Cross-Over for Magnetotelluric (MT) 1D Data Inversion
Author(s) -
Eki Dwi Wijanarko,
Hendra Grandis
Publication year - 2019
Publication title -
iop conference series. earth and environmental science
Language(s) - English
Resource type - Journals
SCImago Journal Rank - 0.179
H-Index - 26
eISSN - 1755-1307
pISSN - 1755-1315
DOI - 10.1088/1755-1315/318/1/012029
Subject(s) - crossover , algorithm , genetic algorithm , population , computer science , binary code , synthetic data , binary number , coding (social sciences) , binary search algorithm , inversion (geology) , magnetotellurics , search algorithm , mathematics , artificial intelligence , statistics , electrical resistivity and conductivity , engineering , machine learning , demography , arithmetic , sociology , paleontology , structural basin , electrical engineering , biology
The magnetotelluric (MT) 1D modelling has been continuously receiving interest due to its effectiveness in obtaining overall subsurface resistivity image of an investigated area. The advances in computational resources allow increasing implementations of non-linear inversion using global search approach, such as Genetic Algorithm (GA). The genetic algorithm adopts the process of natural selection (survival for the fittest) and genetic transformation, i.e. selection, reproduction, mutation and population change, to solve the optimization problem. This paper discusses GA application in MT 1D modelling using binary coding representation with multi-point cross-over, i.e. one for every model parameter. The model parameters are resistivity and thickness of homogenously horizontal layers. The algorithm parameters need to be set to work properly, i.e. population size, number of genes, number of bits, crossover probability, mutation probability and number of generations. Despite binary coded GA (BCGA) drawbacks discussed in the literatures, we found that binary representation allows relatively extensive exploration of the search space. Test using synthetic data from three-layered synthetic models lead to satisfactory results, in terms of synthetic model recovery and data misfit comparable with the noise level contained the synthetic data.