Drilling system
A method can include acquiring drilling performance data for a downhole tool; modeling drilling performance of the downhole tool to generate results; training a machine learning model using the drilling performance data and the results to generate a trained machine learning model; and predicting behavior of the downhole tool using the trained machine learning model.
1. A method comprising:
acquiring, via one or more downhole tool sensors, drilling performance data for a downhole tool;
modeling drilling performance of the downhole tool to generate results;
training a machine learning model using the drilling performance data and the results to generate a trained machine learning model, wherein training the machine learning model includes using the drilling performance data and the results to compute residuals associated with values of a drilling trajectory that comprises a dogleg during drilling;
predicting behavior of the downhole tool using the trained machine learning model;
selecting, based at least on the behavior of the downhole tool, the downhole tool for drilling a borehole;
operating the downhole tool in regularly spaced intervals that are proportioned into neutral periods and bias periods;
cycling a tool face of the downhole tool at a first rate during the neutral periods such that a net trajectory response of the downhole tool is approximately tangent with zero net curvature;
cycling the tool face of the downhole tool at a second rate during the bias periods such that the net trajectory response of the downhole tool has a curvature;
acquiring, via the one or more downhole tool sensors, second drilling performance data for the downhole tool during the drilling of the borehole; and
dynamically adjusting the trained machine learning model in real-time based on the second drilling performance data during the drilling of the borehole.
2. The method of claim 1 , comprising:
outputting, via the trained machine learning model, control instructions based in part on the behavior of the downhole tool; and
controlling the drilling of the borehole based in part on the control instructions.
3. The method of claim 1 , wherein the modeling the drilling performance of the downhole tool comprises utilizing a physics-based model.
4. The method of claim 1 , wherein the machine learning model comprises a Gaussian process model.
5. The method of claim 1 , wherein the trained machine learning model comprises an ensemble model.
6. The method of claim 1 , wherein the training the machine learning model using the drilling performance data and the results to generate the trained machine learning model comprises computing one or more additional residuals.
7. The method of claim 6 , wherein the one or more additional residuals comprise errors between the drilling performance data and the results.
8. The method of claim 1 , comprising rendering a graphical user interface to a display based on the predicting the behavior of the downhole tool, optionally wherein the graphical user interface comprises dogleg information for the downhole tool.
9. The method of claim 1 , comprising providing a trained pre-processing machine learning model that outputs one or more values for the downhole tool for the modeling the drilling performance of the downhole tool.
10. The method of claim 1 , wherein the downhole tool comprises a rotary steerable system.
11. A system comprising:
a processor;
memory accessible by the processor;
processor-executable instructions stored in the memory and executable to instruct the system to:
acquire, via one or more downhole tool sensors, drilling performance data for a downhole tool;
model drilling performance of the downhole tool to generate results;
train a machine learning model using the drilling performance data and the results to generate a trained machine learning model, wherein to train the machine learning model, the processor-executable instructions stored in the memory and executable to further instruct the system to compute residuals associated with values of a drilling trajectory that comprises a dogleg during drilling based on the results and the drilling performance data;
predict behavior of the downhole tool using the trained machine learning model;
select, based at least in part on the behavior of the downhole tool, the downhole tool for drilling a borehole;
generate control instructions for operating the downhole tool in regularly spaced intervals that are proportioned into neutral periods and bias periods;
generate control instructions for cycling a tool face of the downhole tool at a first rate during the neutral periods such that a net trajectory response of the downhole tool is approximately tangent with zero net curvature;
generate control instructions for cycling the tool face of the downhole tool at a second rate during the bias periods such that the net trajectory response of the downhole tool has a curvature;
acquire, via the one or more downhole tool sensors, second drilling performance data for the downhole tool during drilling of the borehole; and
dynamically adjust the trained machine learning model in real-time based on the second drilling performance data during the drilling of the borehole.
12. One or more computer-readable storage media comprising processor-executable instructions to instruct a computing system to perform the method according to claim 1 .
13. A method comprising:
acquiring first drilling performance data for a downhole tool;
modeling drilling performance of the downhole tool to generate results;
training a machine learning model using the first drilling performance data and the results to generate a trained machine learning model, wherein training the machine learning model includes using the drilling performance data and the results to compute residuals associated with values of a drilling trajectory that comprises a dogleg during drilling;
predicting behavior of the downhole tool using the trained machine learning model;
controlling a drilling operation that uses the downhole tool;
acquiring, via one or more downhole tool sensors, second drilling performance data for the downhole tool during the drilling operation;
dynamically adjusting the trained machine learning model in real-time based on the second drilling performance data during the drilling operation;
outputting, via the trained machine learning model, control instructions for the adjusting the downhole tool;
operating, based in part on the control instructions, the downhole tool in regularly spaced intervals that are proportioned into neutral periods and bias periods;
cycling, based in part on the control instructions, a tool face of the downhole tool at a first rate during the neutral periods such that a net trajectory response of the downhole tool is approximately tangent with zero net curvature; and
cycling, based in part on the control instructions, the tool face of the downhole tool at a second rate during the bias periods such that the net trajectory response of the downhole tool has a curvature.
14. The system of claim 11 , wherein the one or more downhole tool sensors include at least one of a first sensor for measuring rotations per minute of a collar of the downhole tool, a second sensor for measuring a steering ratio demand of the downhole tool, or a third sensor for measuring a parameter of the tool face of the downhole tool.
15. The system of claim 11 , wherein to generate the control instructions for drilling the borehole with the downhole tool, the processor-executable instructions stored in the memory and executable to further instruct the system to:
quantize drilling time cycles into regularly spaced intervals, wherein the regularly spaced intervals are further proportioned into the neutral and bias periods.
16. The method of claim 13 , wherein the one or more downhole tool sensors include at least one of a first sensor for measuring rotations per minute of a collar of the downhole tool, a second sensor for measuring a steering ratio demand of the downhole tool, or a third sensor for measuring a parameter of the tool face of the downhole tool.