CPC G01V 5/101 (2013.01) [E21B 49/00 (2013.01); E21B 49/005 (2013.01); G01N 33/24 (2013.01); G06N 3/045 (2023.01); G06N 3/08 (2013.01); G06N 3/084 (2013.01); G06N 3/088 (2013.01); E21B 49/02 (2013.01); E21B 2200/20 (2020.05); E21B 2200/22 (2020.05); G06N 3/082 (2013.01)] | 20 Claims |
1. A method for characterizing a geological formation comprising:
a) generating or obtaining data pertaining to concentrations of a set of atomic elements in a part or sample of the geological formation based on at least one measurement of the part or sample of the geological formation;
b) using the data of a) as input to a mapping function that derives at least one of i) concentrations of a set of mineral components in the part or sample of the geological formation or ii) reconstructed concentrations of the set of atomic elements in the part or sample of the geological formation, wherein the mapping function is based on training a neural network; and
c) determining at least one parameter characterizing the part or sample of the geological formation based on the concentrations of the set of mineral components in the part or sample of the geological formation or the reconstructed concentrations of the set of atomic elements in the part or sample of the geological formation,
wherein the mapping function of b) is derived from minimization of a cost function given a set of data comprising: input data, latent space data, output data, uncertainties in the input data, uncertainties in the latent space data, missing data, and data of different fidelities as captured by their uncertainties.
|