An Adaptive Layering Dual-Parameter Regularization Inversion Method for an Improved Giant Trevally Optimizer Algorithm

Abstract

The inversion effect of the initial model constructed directly from prior information is highly dependent on the quality of the prior information. To reduce the inversion deviation caused by the inaccuracy of the prior information, this paper proposes an Optimized Dual-parameter Regularization (ALS-ODR) inversion method with an Adaptive Layering Strategy. Firstly, a layered model is established based on prior information, and an initial inversion model is constructed by preliminarily selecting the layering values for each segment of the layered model. Secondly, a dual-parameter regularization method is utilized to construct the inversion objective function, addressing the problem of inversion multiplicity caused by the increase in inversion parameters due to the increased layering numbers. Subsequently, the current model parameters of the inversion objective function are optimized using the Giant Trevally Optimizer (GTO) algorithm, improved by the Particle Swarm Optimization (PSO) algorithm. Then, according to the adaptive layering strategy, the model inversion calculation is continuously performed until the optimal inversion model solution is found. The ALS-ODR inversion method is evaluated on one-dimensional and two-dimensional models using different regularization methods, layering methods, and inversion algorithms. The method is also applied to explore the transient electromagnetic field data in a mining area in Chongqing. Simulation experiments demonstrate that the ALS-ODR inversion method improves the clarity of the anomaly boundary and the stability of the inversion results. Additionally, field data experiments also validate that the ALS-ODR inversion method exhibits better practicality and higher fitting accuracy.

Keywords

• Dual-parameter regularization inversion
• adaptive layering strategy
• improved giant trevally optimizer algorithm
• transient electromagnetic exploration