全空间条件下矿井瞬变电磁法粒子群优化反演研究

煤矿井下矿井瞬变电磁法(MTEM)探测中,电磁场呈全空间分布,全空间瞬变电磁反演是复杂的非线性问题,目前反演计算中全空间响应主要由半空间响应乘以全空间响应系数来得到,导致反演结果中顶板和底板异常(或前方和后方异常)叠加在一起难以分离,造成分辨率下降.论文提出采用粒子群优化算法(PSO)进行全空间MTEM反演,通过理论分析,在常规的粒子群算法基础上提出了一种新的进化公式改进策略,提高了粒子群算法的寻优能力.基于全空间瞬变电磁场理论,编写了粒子群算法反演程序,进行全空间条件下五层含巷道的复杂模型的反演计算.结合某矿井巷道顶板、底板岩层及断层含水性的探测实例,对实测数据进行反演计算和解释,探测结果得到钻探证实.研究表明,改进的粒子群优化算法对理论模型和实际资料的反演拟合程度较高,实现了矿井顶板、底板视电阻率异常的分离,提高了全空间瞬变电磁勘探资料的解释精度和分辨率.

Study on particle swarm optimization inversion of mine transient electromagnetic method in whole-space

In underground mining detecting using the mine transient electromagnetic method (MTEM), electromagnetic field is of whole-space distribution. Mine transient electromagnetic inversion in whole-space is nonlinear and complicated. At present, whole-space response is mainly obtained by the product of half-space response and the whole-space response coefficient in conventional inversion, which makes the anomalies of roof and floor (or front and rear) overlap together and difficult to separate, thus the resolution decreases. In this paper, particle swarm optimization (PSO) was adopted for whole-space inversion of MTEM. Based on theoretical analysis and standard particle swarm optimization algorithm, a new evolution formulation was proposed to improve the optimization ability. Particle swarm optimization inversion program was compiled based on whole-space transient electromagnetic field theory, and then the inversions of a complex five-layer model in whole-space were made respectively. Combined with a case of detecting water-bearing zone of mine roadway roof and floor rock mass and fault, the results of inversion calculation and interpretation for the surveyed data were confirmed by drilling. The research showed that the inversion fitted well for theoretical model and the surveyed data respectively by the improved particle swarm algorithm, the separation of the mine roof and floor apparent resistivity anomalies was realized, and the accuracy and resolution of whole-space transient electromagnetic detection were improved.