位场数据群智能随机反演

• 作 者: 刘双 胡祥云 习宇飞 蔡建超 张恒磊 李建慧 韦伟
• 机 构: 中国地质大学(武汉) 武汉 430074 中国地质科学院水文地质环境地质研究所 石家庄 05006
• 出 处: 2015年中国地球科学联合学术年会
• 发表日期: 2015年11期
• 关键词: 地质勘探 位场数据群 智能随机反演技术

Swarm intelligence ant colony optimization(ACO)algorithm is widely applied to solve combinationaloptimization problems(COP),but it also can be utilized to invert for potential field data according tooptimizing the objective function and dividing the continuously model parameters into discrete nodes.Antcolony tour the nodes to find the optimal solution just as the salesman visits all the cities to search the shortestpaths in traveling salesman problem(TSP).This node partition strategy ant colony optimization(NP-ACO)can be used in two kinds of potential field data inversion:parametric inversion of simple shape model andimaging inversion of complicated shape model.And the method achieves good inversion results both insynthetic and real examples.Because it is based on the information positive feedback mechanism,NP-ACO shows better optimizationcapacity.And it has some intrinsic advantages such as steady convergence,strong robustness,good parallelimplementation and convenient portability compared with other stochastic metaheuristics.Both the binaryrepresentation imaging and full imaging ACO can produce sharper physical properties distributions than otherlinear inversion methods such as preconditioned conjugate gradient(PCG).ACO has potential to be apopularinversion approach in geophysical data inversion like Monte Carlo(MC),genetic algorithm(GA),simulatedannealing(SA),artificial neural network(ANN).Nonetheless,because ground potential field data have lower resolution at vertical direction,it is difficultto distinguish the deeper or closed target bodies using this method.NP-ACO also has some drawbacks.Forinstance,it takes too much time to find the shortest route and the ant colony are prone to mature too early,eventhough our proposed Gaussian mapping has greatly promoted the convergence rate.