A Study On Application Of BP-GA Mixture Algorithm In Impedance Inversion

Wave impedance inversion of seismic inversion is a very important one, whichis completely nonlinear inversion of the current research hotspot. The artificial neuralnetworks theory as a new nonlinear science, with neural network researchdevelopment has been widely used in various fields. Genetic algorithm is a simulationnatural evolution of search algorithm, which as part of the evolutionary computation.Because of its space if not for continuous and derivative, parallel processingcapabilities in various fields have been widely used. But traditional BP networks andgenetic algorithms are lacking, BP networks are training a long time, vulnerable to thelimitations of training samples and fault tolerance of difference, genetic algorithmprone to premature phenomena such as inadequate. This paper introduces the artificialneural networks, genetic algorithms theoretical basis, and design BP-GA hybrid ofartificial neural network training algorithms and evolutionary neural networkalgorithm. The BP-GA hybrid algorithm using trigger probability, in the traditional BPalgorithm automatically called genetic algorithms for network optimization weights,and optimizing the GA, such as improving the coding and decoding algorithm, usingthe roulette operator with the option to preserve the optimal combination of optionsstrategy, and using the more approximate Gaussian variation against precocious issues.Traditional forward networks after completed the training network to a fixed value,can't adjust in further application. The evolutionary neural network algorithm can testeach output. When the error exceeds the scope of a given genetic algorithm called onthe network to value adjustments, improve network performance.Using BP algorithm and BP-GA hybrid algorithm inversion theoretical modelwith the different initial model, different wavelet, and different noise. Based on theresults of the comparative analysis expounded the BP-GA hybrid algorithm advantageand the actual data inversion is a good result. Finally, the BP wave impedanceinversion algorithms were discussed, summarize the BP-GA hybrid algorithms,analysis of the limitations of the algorithm, and expound the forward research.