摘要In this paper, we focus on the influences of various parameters in the niching genetic algorithm inversion procedure on the results, such as various objective functions, the number of the models in each subpopulation, and the critical separation radius. The frequency– waveform integration (F–K) method is applied to synthesize three-component waveform data with noise in various epicentral distances and azimuths. Our results show that if we use a zero-th-lag cross-correlation function, then we will obtain the model with a faster convergence and a higher precision than other objective functions. The number of models in each subpopulation has a great influence on the rate of convergence and computation time, suggesting that it should be obtained through tests in practical problems. The critical separation radius should be determined carefully because it directly affects the multiextreme values in the inversion. We also compare the inverted results from full-band waveform data and surfacewave frequency-band (0.02–0.1 Hz) data, and find that the latter is relatively poorer but still has a higher precision, suggesting that surface-wave frequency-band data can also be used to invert for the crustal structure.
Abstract:In this paper, we focus on the influences of various parameters in the niching genetic algorithm inversion procedure on the results, such as various objective functions, the number of the models in each subpopulation, and the critical separation radius. The frequency– waveform integration (F–K) method is applied to synthesize three-component waveform data with noise in various epicentral distances and azimuths. Our results show that if we use a zero-th-lag cross-correlation function, then we will obtain the model with a faster convergence and a higher precision than other objective functions. The number of models in each subpopulation has a great influence on the rate of convergence and computation time, suggesting that it should be obtained through tests in practical problems. The critical separation radius should be determined carefully because it directly affects the multiextreme values in the inversion. We also compare the inverted results from full-band waveform data and surfacewave frequency-band (0.02–0.1 Hz) data, and find that the latter is relatively poorer but still has a higher precision, suggesting that surface-wave frequency-band data can also be used to invert for the crustal structure.
基金资助:the National Natural Science Foundation of China (Nos. 41274059, 40974021 and 40774044) and Beijing Natural Scientific Foundation (Nos. 8122039 and 8092028) to J. Lei
通讯作者:
Jianshe Lei
E-mail: jshlei_cj@hotmail.com
引用本文:
Cong Li, Jianshe Lei. Numerical tests for effects of various parameters in niching genetic algorithm applied to regional waveform inversion[J]. 《地震学报》英文版, 2014, 27(5): 541-551.
Cong Li, Jianshe Lei. Numerical tests for effects of various parameters in niching genetic algorithm applied to regional waveform inversion. Earthquake Science, 2014, 27(5): 541-551.