A Comparative Analysis of AI Models and Agents for Diagnosing and Mitigating Sloughing Shale Issues in Deep Oil Wells: A Case Study of a 20,000-Foot Well

Abstract

This scientific paper introduces an applied study using 16 artificial intelligence (AI) models and agents to manage a technical problem encountered while drilling a 20,000 ft oil well. The issue was sloughing shale, where formation layers collapse into the wellbore. The AI systems were trained on a comprehensive dataset of images, texts, videos, & numerical data, with sensitive details like well location and company name withheld for confidentiality. The AI identified and diagnosed the problem, proposed effective treatments, and recommended preventative measures. Furthermore, it was required to design a suitable drilling fluid for the problematic depth and suggest optimal drilling parameters to prevent future occurrences. The results revealed a significant variance in performance. AI agents such as Minmax and Manus demonstrated 100% accuracy in diagnosing the problem and proposing both remedial and preventive solutions. In contrast, AI models like GPT and Gemini achieved 90-95% accuracy, while the remaining models performed at approximately 80%. These findings underscore the immense potential of AI in enhancing oil drilling operations, from early problem detection to designing technical solutions. The study also highlights AI’s vital role in training personnel by simulating operational challenges. It concludes by emphasizing the importance of integrating advanced AI agents into oil and gas workflows to boost operational efficiency, mitigate risks, and lower costs, thereby accelerating the sector's digital transformation.