3D resistivity inversion using an improved Genetic Algorithm based on control method of mutation direction

摘要: Abstract Traditional inversion method is the most commonly used procedure for three-dimensional (3D) resistivity inversion, which usually takes linearization of problem and accomplish it by iterations. However, its accuracy often dependent on initial model, can make trapped in local optima, even cause a bad result. Non-linear feasible way to eliminate dependence model. large problems such as 3D with parameters exceeding thousand, main challenges non-linear are premature quite low search efficiency. To deal these problems, we present an improved Genetic Algorithm (GA) method. In GA method, smooth constraint inequality both applied object function, degree non-uniqueness ill-conditioning decreased. Some measures adopted from others reference maintain diversity stability GA, e.g. real-coded adaptive adjustment crossover mutation probabilities. Then generation approximately uniform population proposed this paper, uniformly distributed be produced model eliminated. Further, direction control presented based joint algorithm, embedded GA. The update vector increment better compared traditional non-controlled operation. By optimized efficiency greatly. performance evaluated comparing results synthetic example or drilling columnar sections practical example. examples illustrate that obtain high-quality results.