A Priori Constrained ACO Method Applied to Three-Dimensional Imaging of Subsurface Electrical Resistivity

摘要

Electrical resistivity imaging seeks to reconstruct conductivity model of subsurface water-bearing body for disaster reduction and resource protection. Since its high non-linearity, the inverse problem relying mainly on linear least-square method may inclined to consider much deeper adoption of non-linear algorithm. Ant colony optimization (ACO) is a standard probabilistic algorithm used to find optimal paths. Since the direct application of the ACO is unstable, there are many false anomalies, and prior information needs to be applied according to the laws of physical properties. This study proposed a priori constrained approach to improving ant colony optimization method wherein prior information is included as regularization term in search process. First, we introduce foraging path smoothness into objective function to reduce non-uniqueness. On this basis, smooth constraint of resistivity is added into visibility and inequality constraint is added into pheromone intensity, by which some unnecessary nodes of foraging paths are avoided form selecting by ants. Finally, the strategy of global minimum is introduced in the process of updating pheromone. Forward modeling of each ant selected is parallel calculated via shared-memory computing strategy. Taking into account of those, the direction of global minimum is maintained and the search efficiency is improved greatly, which makes the constrained ACO feasible for the 3-D resistivity inversion. Both numerical experiments and field application is conducted to evaluate this method and illustrate that we can obtain the improved inversion results of subsurface water-bearing body.