The Assisted Cement Log Interpretation project has used machine learning (ML) to create a tool that interprets cement logs by predicting a predefined set of annular condition codes used in the cement log interpretation process. The development of a cement log interpretation tool speeds up the log interpretation process and enables expert knowledge to be efficiently shared when training new professionals in the cased hole logging unit. By using high quality and consistent training data sets, the project has trained a model that will support unbiased and consistent interpretations over time. The tool consists of a training and a prediction tool integrated with the cased hole logging interpretation software. By containerizing the code using an "API First" design principle, the applicability of this add- on tool is broad. The ML model is trained using selected and engineered features from cement logs, and the tool predicts an annular condition code according to the cement classification system for each depth segment in the log. The interpreters can easily fetch a complete cement log interpretation prediction for the log and use that as a template for their final interpretation. The ML model can easily be retrained with new data sets to improve accuracy even further. To improve cement log interpretation consistency in the industry, the results are made available as open source.

The Assisted Cement Log Machine Learning (ML) tool – or Annular barrier AI tool - developed by Equinor - is being used to interpret cement logs by predicting a predefined set of annular condition codes used in the cement log interpretation process. (Reference to SPE paper: Assisted Cement Interpretation project). Annular conditions are usually separated into High, Medium and Low probability for hydraulic isolation. The internally developed annular condition code descriptions at Equinor are separated into 30 specific classes, which supports more nuanced and objective expert interpretations. The paper will discuss how this framework has positively impacted the performance of the trained ML model. In addition, we report how this tool is being used to speed up and increase consistency in the log interpretation process, and how it can be used to efficiently share expert knowledge when training new professionals into Equinor's Cased Hole Logging Group. Furthermore, the paper will discuss ongoing research to improve the capabilities of this tool, like supporting the use of cement logs from additional service vendors, and how it could be potentially expanded to extract relevant information from historical reports to improve formation bond predictions. The ML model is trained using selected and calculated features from cement logs, and the tool predicts an annular condition code according to the cement classification system for each depth segment in the log. Training and prediction are done in the cloud and accessible through an Application Programmable Interface (API) which makes it convenient to integrate the tool with any cement log interpretation software. Through the API, the interpretation software uploads a cement log and swiftly receives predictions for the complete log, including hydraulic isolation probabilities and confidence curves, which are used as decision support for the final expert interpretation. The ML model is regularly retrained with an ever-growing data set from real operations performed by Equinor. The uploaded data undergoes an automatic quality assurance before it is used as training data, and the model's performance is evaluated at each retraining. To improve the cement log interpretation consistency in the industry and to ensure that our work can benefit the industry as widely as possible, the results will be made available as open source. This paper will discuss the challenges making such an ML tool open source, and the how the idea of Federated Learning could be used to share this solution in the industry.

Summary The Assisted Cement Log Interpretation Project has used machine learning (ML) to create a tool that interprets cement logs by predicting a predefined set of annular condition codes used in the cement log interpretation process. The development of a cement log interpretation tool speeds up the log interpretation process and enables expert knowledge to be efficiently shared when training new professionals. By using high-quality and consistent training data sets, the project has trained a model that will support unbiased and consistent interpretations over time. The tool consists of a training and a prediction tool integrated with cased-hole logging interpretation software. By containerizing the code using an “API First” design principle (API: application programming interface), the applicability of this add-on tool is broad. The ML model is trained using selected and engineered features from cement logs, and the tool predicts an annular condition code according to the cement classification system for each depth segment in the log. The interpreters can easily fetch a complete cement log interpretation prediction for the log and use that as a template for their final interpretation. The ML model can easily be retrained with new data sets to improve accuracy even further. To improve cement log interpretation consistency in the industry, the code will be made available as open source.