Application of K-means algorithm to Werner deconvolution solutions for depth and image estimations

One of the most popular techniques for computer-assisted solution estimates for magnetics and gravity field data is Werner deconvolution. The approaches frequently produce erratic results and may not always forecast the maximum number of the geologic entity that produces them due to the intrinsic instability of potential field data. This led to the application of the K-means machine learning algorithm to further enhance the detection of the geologic potential field-generated bodies. Two substances that resembled dikes were combined to form a synthetic magnetic model. Random noise was added to the synthetic data, to make the solutions a bit more complex. Werner deconvolution technique was applied to the synthetic model to generate solutions. K-means unsupervised machine learning algorithm was applied to the generated solutions created by the synthetic data. We further applied this algorithm to real data sets from a mining site. The clustering result shows a good spatial correspondence with the geologic model, and the method was able to estimate the precise location and depth of the dike bodies. The proposed method is entirely data-driven and has proven to work in the presence of noise.