Can drillers extract insights from successful and challenging cases by writing one sentence? Today, the drillers either dig, for days or weeks, the mixed-structured data of the Daily Drilling Report (DDR), the structured drilling data, or both to extract knowledge about successful cases (e.g., a record rate of penetration) and challenging cases (e.g., stuck pipe and Non-Productive Time (NPT)). The objective is to have the drilling operations insights extracted with no time from the current and historical data reports.
We propose a more efficient knowledge extraction of drilling operations in seconds or minutes by writing one sentence using the latest artificial intelligent Chat Generative Pretrained Transformer algorithm (ChatGPT). Therefore, we launched the first drilling dedicated ChatGPT pilot. ChatGPT has pretrained models; however, in this pilot, we enable ChatGPT to learn from our drilling data to provide specific answers to our challenges accurately and efficiently. The implementation method of ChatGPT requires multiple stages: (1) Data Loading/Downloading and Document Scanning, (3) Data Indexing, (4) ChatGPT Training, and (5) ChatGPT extraction of knowledge.
Our drilling data is available in structured (tabulated), unstructured, and mix-structure formats; therefore, understanding the behavior of ChatGPT in these different formats and other training indexing and cognitive capabilities are some of the pilot targeted objectives.
This novel pilot is the first in the oil industry to use ChatGPT, particularly in drilling. Its outcome determines ChatGPT's ability to ease drilling operations by providing insight and learning from historical success and challenging cases. This paper reveals the methods and tools to quickly deliver efficient and quality answers to drilling operations to the drilling engineers.