NLP Model for Drilling Event Detection in Well Reports
Find Innovative SolutionsGenerate Solutions
Solution Overview
Problem
The free-text component of drilling reports is challenging to access effectively due to differences in word-choice, making it difficult to utilize valuable insights from these reports for subsequent well drilling operations.
Innovation Solution
A computer-implemented method for identifying drilling events and activities from text data in drilling reports using a model, which includes receiving drilling reports, identifying events and activities, obtaining feedback, training the model, and creating or confirming well plans based on the identified events and activities.
Engineering Contradictions & Design Principles
Engineering Contradiction Analysis
1Loss of information
If free text is used in drilling reports to capture detailed observations, then the quantity and depth of information increase, but the difficulty of accessing and analyzing this information increases due to word-choice variations
Solution Approach 1:
A natural language processing system acts as an intermediary between the unstructured free text in drilling reports and the structured data needed for analysis. The NLP model translates varied word choices and unstructured narratives into standardized event classifications and structured data fields, enabling consistent access to information while hiding the complexity of text variation from the analysis process.
Solution Approach 2:
The system transforms the parameter representation of text data from unstructured natural language with variable word choices into structured parameters with standardized categories and values. By changing the parameter space from free-text strings to coded event types and standardized attributes, the system maintains information richness while enabling systematic access and analysis.
2Loss of information
If manual analysis of free-text reports is performed, then detailed insights can be extracted, but the time and resources required increase significantly
Solution Approach 1:
The manual mechanical process of reading and analyzing free-text reports is replaced with an automated computational system using natural language processing. The NLP model automatically processes text data, identifies drilling events, extracts entities, and classifies observations without human intervention, maintaining insight quality while dramatically reducing the time and human resources required.
Solution Approach 2:
The system enables self-service analysis where the drilling reports automatically process and analyze their own free-text content through the NLP pipeline. The model autonomously identifies events, extracts entities, and structures information without requiring manual analysis, allowing the data to serve itself and eliminate the time-consuming manual review process.
3Loss of information
If comprehensive free-text descriptions are recorded during drilling, then more contextual information is captured, but the difficulty of standardizing and comparing data across reports increases
Solution Approach 1:
The NLP system serves as an intermediary layer that receives comprehensive free-text descriptions with all their contextual nuances and transforms them into standardized data representations. The model preserves contextual information by understanding semantic relationships in the text while outputting standardized event types, entities, and attributes that enable consistent comparison across different reports.
Solution Approach 2:
The system performs a parameter transformation from unstandardized natural language parameters to standardized data parameters. By mapping varied text descriptions to a controlled vocabulary of event types and standardized attribute schemas, the system maintains the rich contextual information from comprehensive descriptions while achieving the standardization needed for consistent data comparison and analysis.
Data Source
Figure 1
Figure 2
Figure 3
AI summary
A system, computer-readable medium, and method for identifying drilling events in drilling reports, of which the method includes receiving one or more drilling reports including text data representing one or more observations recorded during a drilling activity, identifying a drilling event, a drilling activity, or both from the text data using a model, obtaining feedback based at least in part on the drilling event, the drilling activity, or both that were identified, and training the model based on the feedback.