GRAVITY INVERSION MODELLING USING SIMULATED ANNEALING AND THE LEVENBERG-MARQUARDT ALGORITHM

Abstract

The gravity method is a geophysical method that can model subsurface conditions based on variations in density. This model can be obtained by performing a mathematical computation process of gravity data which is known as the inversion modeling process. During its development, inversion modeling solutions can be obtained with local and global approaches. A comparison of the two approaches was carried out, namely using the Levenberg-Marquardt (LM) algorithm for the local approach and the simulated annealing (SA) algorithm for the global approach. The research begins with the study of synthetic data horizontal block, vertical block, and faults models. From the synthetic study, both algorithms were rated good based on an average root mean square error (RMSE) value of 0.03 mGal with a correlation of 76% for the LM algorithm and an average RMSE of 0.6 mGal with a correlation of 61% for the SA algorithm. The two algorithms were also tested on the Ungaran geothermal field gravity data. RMSE value was 0.001 mGal for the LM algorithm and 0.7 mGal for the SA algorithm due to the field data test. The two models produced by each algorithm are assessed to be correlated with geological information and models in previous studies.