Minimizing Unplanned Maintenance Shutdowns

Optimizing Oil and Gas Acid Gas Removal Units (AGRU) is crucial, but breakthroughs may cause CO2 residuals, leading to unplanned shutdowns. We responded with advanced predictive modeling and monitoring tools that offered proactive insights to prevent disruptions and ensure continuous production by addressing breakthrough-related issues promptly.

Optimizing Dataset Creation and Hypothesis Testing

To handle the vast amount of sensor data from plant sensors, we leveraged advanced resampling techniques to meticulously tailor a dataset, ensuring an optimal frequency for thorough analysis. Additionally, through rigorous hypothesis testing, we arrived at an ideal alert interval, optimizing responsiveness within the 10–30-minute range.

Predictive Modeling for Enhanced Accuracy

Challenged by the scarcity of data for the class, we implemented a groundbreaking solution by introducing synthetic data for event training, significantly elevating the overall prediction accuracy of our model. Further enhancing precision, we employed a probability optimization algorithm during training, substantially reducing false alarms. To seamlessly integrate this advanced predictive model into the native language of the Advance Process Control System, we designed a dedicated conversion pipeline. This not only ensured compatibility but also improved operational efficiency.

Empowering Engineers with a Model Monitoring UI Tool

To provide actionable insights, we introduced an intuitive user interface. This user-friendly layer empowered production engineers to conduct in-depth analyses of crucial metrics, facilitating a deeper understanding of events and identifying key drivers behind them. The Model Monitoring UI Tool emerged as a crucial component, empowering our team with real-time insights and fostering a proactive approach to operational management.