Inversion of geological and geophysical data sets using genetic algorithms

A technique is presented for the multiply-constrained inversion of geological and potential field data. This technique speeds up the process of testing simplified kinematic structural models of structural and potential-field data collected by field geologists and geophysicists. The scheme uses pre-existing software (Noddy) to calculate possible solutions, and as the inverse modeling scheme requires flexibility and speed over a potentially large parameter space with many local minima, a genetic programming approach to the global optimization problem is used. The Noddy integrated forward modeling package calculates geometries and potential-field anomalies resulting from a specified geologic structure. These models are then compared to the target data using correlation of potential-field anomaly images to assign the fitness of the model, with the genetic algorithm breeding successive generations of models. Initial testing of the inversion scheme was successful for a spherical plug, and a planar dike.