| Popis: |
A dynamic digital twin solution has been implemented to enable process equipment condition monitoring on key topsides equipment such as gas compressors, compressor suction and discharge coolers, seawater lift pumps, cooling medium coolers, and various filters. Equipment condition monitoring is enabled by performing virtual measurements, equipment degradation assessment and detection of anomalous process conditions. The existing Stampede Multi-Purpose Dynamic Simulation model was repurposed as a real-time performance monitoring model. This model was connected to a historian to read real-time data enabling virtual measurement and equipment performance analysis. Dynamic online clustering of the process variables measured from field and rolling regression on the performance monitoring model output is performed to estimate the remaining useful life and detect anomalous process conditions. Virtual measurements from the real-time digital twin gives more insight into the performance of the production facility, which helps operators and engineers more efficiently troubleshoot equipment issues. The proposed data analytics methodology is implemented on a Boost Gas Compressor (BGC) installed on the Stampede facility. Two years of historic data from the Stampede facility has been used for this case study to estimate remaining useful life and detect anomalous operation conditions that result in damage to the system eventually requiring a replacement or causing a shutdown. For the BGC case study presented, we were able to detect the anomalous operating conditions for the BGC about 6 days prior to a defect being detected by the maintenance crew on the discharge cooler. It is possible to avoid damage to the gas compression system if we can detect when the BGC is being operated in an anomalous operating condition and prevent early replacement of expensive equipment. The proposed dynamic digital twin provides better insights for condition monitoring of equipment in an oil and gas facility using both the field sensor data and the physics-based models. In addition to performance deviation measurements, integration of field sensor data with physics-based models provides virtual measurement at locations where actual field sensors are not available. Dynamic clustering helps tackle the lack of information about gas molecular weight in real-time. |
