The Research And Application In Velocity Inversion Based On Rapid Genetic Algorithm

Genetic Algorithm(GA) is a fast-developing in recent years and novel nonlinear inverse method,it has strong adaptability and numerical calculation stability,and it is different from other existing inversion methods.GA is a kind of strong practicable inversion method and has all the advantages of inversion directly.Using a model group instead of a model value to search in parallel in the model space makes GA have a better global search capability to reduce the risk of inmeshing into local optimal solution,using general information of object function instead of caculating derivative or other special assistant information,with no harsh hypothesis and restrictions,and using random transfer rules makes Genetic algorithm become a strong Global optimization method of nonlinear inversion.Genetic algorithm has three basic steps:the selection and reproduction,crosscover,mutation.Each step plays individual advantage and makes GA operate normally.Selection and reproduction are based on the value of the objective function to measure the fitness of individuals,and in accordance with the fitness value decide to which ones are copied.Crossover is a way of global search in the model parameters space,it makes a structured exchange of information.Mutation enables groups to maintain a certain degree of variations and random messages,so that it can introduce new genetic materials and informations. Three steps in the genetic algorithm play different roles,and they are very important.This paper introduces the structure and steps of standard genetic algorithm(SGA) ,according to advantages and limitations of SGA,emphatically introduces a rapid genetic algorithm(RAGA),which improves SGA through raising the computation to more efficient. RAGA has better practicability in velocity inversion. RAGA keeps the basic structure and computing steps of SGA,and produces one or two iterations in each rapid loop and optimizes the searching range by using the boundary value of some excellent individuals, so the computation efficiency of the inversion algorithm is greatly improved.In each iteration,RAGA uses probability distribution instead of traditionary fitness calculation so that it can choose excellent individuals easier.Besides,RAGA introduces the update probability,comparing the objective function values of individuals after crossover and mutation with that before crossover and mutation,and it can choose better individuals again to decide real offsprings,these measures improve both calculative speed and convergent speed.After the iteration, we consider that the best individual has been very close to true value,so a chaotic optimization is applied to increase the accuracy and improve the inversion results.The last two chapters of this paper introduce examples of test functions simulation,the simple and complex theoretical velocity model inversion,two-dimensional velocity model inversion and the inversion of actual data,it indicates that RAGA is superior to SGA through runtime analysis and comparison of effects.Not only saving time,improving efficiency,but also improving the accuracy and effictiveness,RAGA is a strong inversion method of high adaptability??practicality.