ROP Projections via Multi-Model Drilling Control
Find Innovative SolutionsGenerate Solutions
Solution Overview
Problem
Existing well drilling technologies face inaccuracies in predicting the rate of penetration (ROP) of drill bits, leading to poor drilling performance, damage to drill bits and wellbores, and operational issues due to unknown or inaccurate lithology of subterranean formations.
Innovation Solution
A system utilizing surface and downhole sensors to collect data, which is then processed by machine-learning models to generate accurate ROP projections, allowing for real-time adjustments to drilling parameters such as revolution-per-minute, weight-on-bit, and mud weight to optimize drill bit performance.
Engineering Contradictions & Design Principles
Engineering Contradiction Analysis
1Measurement precision
If traditional single-model ROP projection methods are used, then the system complexity is low, but the measurement precision of ROP prediction deteriorates
Solution Approach 1:
The patent divides the ROP projection system into multiple independent machine learning models, each specializing in predicting ROP from different data types (lithology data, drilling parameters, real-time measurements). This segmentation allows each model to focus on specific aspects of ROP prediction, improving overall accuracy while maintaining manageable complexity through modular architecture
Solution Approach 2:
The patent combines outputs from multiple specialized machine learning models into a unified ROP projection system. By merging the predictions from lithology-based models, drilling-parameter-based models, and real-time measurement-based models, the system achieves higher measurement precision than any single model could provide alone
2Reliability
If ROP is not controlled accurately, then the drilling speed is high, but the reliability of drilling operations deteriorates
Solution Approach 1:
The patent implements a feedback control system where multiple machine learning models continuously predict ROP based on real-time drilling data, and these predictions are used to adjust drilling parameters. This closed-loop feedback ensures reliable ROP control while maintaining high drilling speed through dynamic parameter optimization
Solution Approach 2:
The system performs preliminary ROP projections using machine learning models before actual drilling operations commence or before entering new lithological zones. This advance prediction allows operators to pre-adjust drilling parameters to achieve optimal ROP, ensuring both reliability and productivity from the start of each drilling segment
3Measurement precision
If multiple machine learning models are used for ROP projection, then the measurement precision improves, but the loss of computational time increases
Solution Approach 1:
The patent segments the computational workload across multiple specialized machine learning models that process different data types in parallel. This segmentation enables simultaneous computation of lithology-based predictions, drilling-parameter-based predictions, and real-time measurements, reducing total computational time while maintaining high accuracy through combined results
Data Source
AI summary
Multiple projected rate of penetration (ROP) values can be determined for purposes of adjusting well tools and well characteristics. For example, surface data can be determined based on a surface sensor signal. Downhole data can be determined based on a downhole sensor signal. A first value indicating a first projected ROP of a drill bit can be determined by providing the surface data as input to a first machine-learning model. A second value indicating a second projected ROP of the drill bit can be determined by providing the downhole data as input to a second machine-learning model. A third value indicating a third projected ROP of the drill bit can be determined by providing the first value and the second value input to a third machine-learning model. An operating characteristic of a well tool can be adjusted based on the third value.