Inversion Of Gravity Anomaly Interface Depth Based On Equilateral Triangle Subdivision And Its Parallel Computing Implementation

Interface depth inversion is a common method for quantitative geological interpretation of gravity anomalies.The geometric shape of the underground density interface can be regarded as a piecewise continuous smooth surface.Based on this,we propose a undulating interface model based on regular triangles.The model has good continuity.Under actual geological conditions,the gravity anomaly in the gravity survey area is not entirely caused by the geological structure in the same size area directly below the survey area.In view of this objective situation,we conducted an edge expansion process on the interface model and gave the empirical principles for determining the epitaxial width and interface depth of the flared part are given.In order to improve the calculation efficiency,we introduced a parallel computing technology based on GPGPU in the inversion process.In the algorithm of in-depth interface inversion,we compare the inversion effect of Bott method and Gauss-Newton method.Finally,we tested the algorithm and program through a synthetic model.Model tests show that: 1)The Bott's method and the Gauss-Newton method can show the undulating shape of the actual density interface.In the error analysis with the actual interface,the interface depth inversion error of the Bott method is basically the same and the whole is shallower than the actual interface.The Gauss-Newton method differs greatly in the error of different measuring points,mainly because there is oscillation at the boundary,but the inversion results at the undulating interface are relatively well fitted to the actual terrain.2)Both inversion methods will improve efficiency with parallel computing.Because the Gauss-Newton method needs to solve large inverse matrices,the Bott method based on linear iteration has higher room for improvement.3)The edge expansion strategy can effectively suppress the jitter and mutation at the boundary,so that the inversion results in the measurement area can be closer to the actual terrain.4)Edge expansion will greatly increase the required computational cost,and at the same time,the effect will decay with the increase of the expansion boundary.Therefore,in actual data processing,it is necessary to select the edge expansion distance according to the average depth of the interface.