世界核地质科学2023，Vol.40Issue(4)：940-951,1008,13.DOI:10.3969/j.issn.1672-0636.2023.04.005
近场源多频电磁法人工蜂群约束反演研究
Study on Constrained Artificial bee Colony Inversion of Near Source Multi-Frequency Electromagnetic Data
摘要
Abstract
Near source multi-frequency electromagnetic method is often used to investigate shallow electrical structures such as engineering geological exploration,groundwater exploration,environmental assessment and underground or underwater buried target detection,etc.Because the near source electromagnetic field cannot be approximated as the plane wave field,one-dimensional inversion is still the most popular method in practical application due to the complexity of two-and three-dimensional forward and inversion.The conventional least square inversion often occurs the non-convergence problem,and its results may be inconsistent with the actual geological condition,which will directly affect the effectiveness of the detection.In order to suppress the negative impacts of convergence instability and strong non-uniqueness of conventional least squares inversion,an inversion method which combines artificial bee colony nonlinear algorithm and prior information constraints was proposed in this paper.Firstly,a brief theoretical introduction of the layered earth forward modeling of near source multi-frequency electromagnetic method was described,and the numerical accuracy of forward modeling was verified.Secondly,the principle of artificial bee colony nonlinear optimization theory was analyzed.The definitions of nectar source and species group of bees were given.Three different kinds of bees including leaders,followers and scouters act in a random way but their group hunting behavior is determined in the sense of complete nonlinear optimization algorithm.Then this algorithm was adopted in the problem of near source multi-frequency electromagnetic data inversion.The objective function was presented,which concerns with the quality of nectar source described in the artificial bee colony algorithm above.Meanwhile,the corresponding work flow of inversion calculation was summarized.After that,the methods of reference model constraint,smooth constraint and physical property range constraint were incorporated into artificial bee colony inversion,and finally the artificial bee colony constrained inversion method of near source multi-frequency electromagnetic method was formed.Compared with the conventional method,it is easier to introduce constraints into artificial colony inversion.The general iteration monotonicity will not be affected by the constraints.The whole numerical algorithm implementation requires almost no changes,except for the objective function interface function and the model update interface function.This integrated approach is expected to enhance the iterative inversion converging procedure while at the same time tending to obtain reliable results guided by constraints.Some typical synthetic models were used to test the operations and the performance of this method,and the results showed that the artificial bee colony inversion method is suitable for multi-layer model,can meet the requirements of observation data fitting,and the target convergence curve decreases steadily as the iterative process continued successively,avoiding the problem of local minimal solution of conventional least squares inversion algorithm.Artificial bee colony inversion was prone to false construction when it was unconstrained,but can effectively tighten the inversion solution space when it was constrained,so that the results were closer to the real value and the quality of the inversion results was better.Two field data cases of near source multi-frequency electromagnetic method were calculated.Their results of the artificial bee constrained inversion were consistent with the actual geological conditions.The results showed that the proposed method is effective and reliable.Therefore,it is recommended to use artificial bee colony constrained inversion method in practice.关键词
近场源多频电磁法/反演/人工蜂群算法/先验信息约束Key words
near source multi-frequency electromagnetic method/inversion/artificial bee colony algorithm/prior information constraint分类
天文与地球科学引用本文复制引用
周俊杰,胡英才,喻翔,陈聪,刘祜..近场源多频电磁法人工蜂群约束反演研究[J].世界核地质科学,2023,40(4):940-951,1008,13.基金项目
核工业北京地质研究院遥感信息与图像分析技术国家级重点实验室基金项目(编号:物ZJ2206)资助. （编号:物ZJ2206）
