US 11,676,000 B2
Drill bit repair type prediction using machine learning
Ajay Pratap Singh, Houston, TX (US); Roxana Nielsen, Conroe, TX (US); Satyam Priyadarshy, Herdon, VA (US); Ashwani Dev, Katy, TX (US); Geetha Gopakumar Nair, Katy, TX (US); and Suresh Venugopal, Spring, TX (US)
Assigned to Halliburton Energy Services, Inc., Houston, TX (US)
Appl. No. 16/611,817
Filed by HALLIBURTON ENERGY SERVICES, INC., Houston, TX (US)
PCT Filed Aug. 31, 2018, PCT No. PCT/US2018/049065
§ 371(c)(1), (2) Date Nov. 7, 2019,
PCT Pub. No. WO2020/046366, PCT Pub. Date Mar. 5, 2020.
Prior Publication US 2020/0149354 A1, May 14, 2020
Int. Cl. G06N 3/042 (2023.01); E21B 10/62 (2006.01); E21B 12/02 (2006.01); G06N 3/08 (2023.01); E21B 10/46 (2006.01); G06N 3/02 (2006.01); G06N 20/00 (2019.01); G06F 30/27 (2020.01); E21B 4/06 (2006.01)
CPC G06N 3/042 (2023.01) [E21B 10/46 (2013.01); E21B 10/62 (2013.01); E21B 12/02 (2013.01); G06F 30/27 (2020.01); G06N 3/02 (2013.01); G06N 3/08 (2013.01); G06N 20/00 (2019.01); E21B 4/06 (2013.01); E21B 2200/20 (2020.05); E21B 2200/22 (2020.05)] 20 Claims
OG exemplary drawing
 
1. A method, comprising:
obtaining drill bit data from a plurality of data sources through one or more application programming interfaces communicably coupled to a processor circuit, the drill bit data including formation lithology information indicating a formation lithology, a drill bit design and a repair history for the formation lithology including a historical distribution of repairs made to individual cutter positions on drill bits that have features of the drill bit design and that have drilled into formations of the formation lithology;
integrating, in a data integration engine executed on the processor circuit, the drill bit data from each of the plurality of data sources into an integrated dataset;
pre-processing, in a data pre-process engine executed on the processor circuit, the integrated dataset to filter out one or more outlier data points from the integrated dataset;
processing, in the processor circuit, the filtered dataset with a neural network to build a machine learning based model;
processing, in the processor circuit, the machine learning based model to extract one or more features that indicate significant parameters affecting wear on a drill bit;
determining, before drilling a wellbore, a repair type prediction with the machine learning based model based on the extracted one or more features, the repair type prediction indicating a predicted type of repair action for the individual cutter positions on the drill bit if drilled into a formation that has the formation lithology;
generating a repair schedule for performing one or more repair actions for the individual cutter positions based on the predicted type of repair actions;
drilling at least a portion of the wellbore into the formation that has the formation lithology; and
providing, after the drilling, a signal indicating a value of the repair type prediction for facilitating a drill bit operation on a cutter of the drill bit based on the repair type prediction.