NLP-Based Fault Detection Method for Multifunction Logging-While-Drilling Services

This paper presents a Natural Language Processing (NLP) method aimed at detecting faults within field failure reports of drilling tools. It builds on the definition of entities specifically matched to our unique requirements. These entities have been annotated within the dataset under the guidance of a Subject Matter Expert (SME), laying a foundation for our NLP method. By utilizing a model based on bidirectional encoder representations from transformers, the method achieves an F1-score of 88\% in identifying entities and consequently detecting faults within field failure reports. This work is part of a long-term project aiming to construct a failure analysis and resolution system for drilling tools.