An AI approach to using magnetic gradient tensor analysis for quick depth and property estimation. Issue 1 (1st December 2019)

• Record Type: Journal Article
• Authors: Pratt, David A.; McKenzie, K. Blair; White, Anthony S.
• Is Part Of: ASEG Extended Abstracts (Online). Volume 2019:Issue 1(2019)
• Page Start: 1
• Page End: 7
• Publication Date: 2019-12-01
• Subjects: magnetic -- tensor -- depth -- AI -- magnetisation; Prospecting -- Geophysical methods -- Periodicals; Prospecting -- Geophysical methods; Periodicals
• DOI: 10.1080/22020586.2019.12073003
• Languages: English
• ISSNs: 2202-0586

Abstract

Summary: Eigenvector analysis of the magnetic gradient tensor is combined with an artificial intelligence (AI) approach to rapid, interactive estimation of depth to magnetic source for a variety of geological target shapes. The method uses the flight line data for maximum depth resolution and grids of the magnetic gradient tensor and other parameters for 2D spatial attributes. The magnetic gradient tensor and related parameters are computed using FFT processing of the original total magnetic intensity grid. These data are then used as input to the pre-trained AI process for preliminary calculation of depth, width and magnetic susceptibility. The eigenvectors are used to compute the normalised source strength (NSS) which peaks over the centre of magnetisation of the magnetic target. The tensor is used to compute the dimensionality of the target which is then used to infer if it is pipe-like or linear. If the target is linear or elongate, the eigenvector analysis provides a direct method for calculating the azimuth of the target at the centre of magnetisation. The azimuth is then used to correct the apparent depth, width and susceptibility estimates. If the target is pipe-like or an ellipsoid in shape, then the eigenvector is used to compute the azimuth and dip of the magnetisation vector. The NSS results also provide a useful tool for estimating the level of interference between adjacent magnetic anomalies, a factor that decreases the accuracy of any magnetic depth estimate. The AI algorithm uses this information to assign a quality estimate to the depth result. At this point, the AI algorithm has derived a lot of information about the target shape, orientation and approximate depth. This information is then used to constrain some classic depth interpretation techniques that include the tensor, Euler 2D, Euler 3D, Peters Length, Werner Deconvolution and Tilt methods. The numerical complexity of each of these methods is greatly simplified because the origin of the target is the centre of magnetisation. Each method has strengths and weaknesses and the AI algorithm attempts to select the best method and most probable geological shape. Interpreters can override both the method and target shape if they are not satisfied with the AI selection because the shape selection has a large influence on the depth estimation precision.