A Modified CGM Algorithm Enhances Data Availability While Retaining iCGM Performance

The algorithm for the Dexcom G6 CGM System was enhanced to retain accuracy while reducing the frequency and duration of sensor error. The new algorithm was evaluated by post-processing raw signals collected from G6 pivotal trials (NCT02880267) and by assessing the difference in data availability after a limited, real-world launch. Accuracy was comparable with the new algorithm—the overall %20/20 was 91.7% before and 91.8% after the algorithm modification; MARD was unchanged. The mean data gap due to sensor error nearly halved and total time spent in sensor error decreased by 59%. A limited field launch showed similar results, with a 43% decrease in total time spent in sensor error. Increased data availability may improve patient experience and CGM data integration into insulin delivery systems.

Measurement-While-Drilling MWD Error Model Validation – Does the Model Reflect Reality?

It has been well documented that magnetic models and Measurement-while-Drilling (MWD) directional sensors are not free from error. It is for this reason that directional surveys are accompanied by an error model that is used to generate an ellipse of uncertainty (EOU). The directional surveys represent the highest probable position of the wellbore and the EOU is meant to encompass all of the possible wellbore positions to a defined uncertainty level. The wellbore position along with the individual errors are typically presumed to follow a Normal (Gaussian) Distribution. In order for this assumption to be accurate, 68.3% of magnetic model and directional sensor error should fall within plus or minus one standard deviation (1σ), 95.5% within two standard deviations (2σ), and 99.7% within three standard deviations (3σ) of the limits defined in the error model. It is the purpose of this study to evaluate the validity of these assumptions. The Industry Steering Committee on Wellbore Survey Accuracy (ISCWSA) provides a set of MWD error models that are widely accepted as the industry standard for use in wellbore surveying. The error models are comprised of the known magnetic model and MWD directional sensor error sources and associated limits. It is the purpose of this paper to determine whether the limits defined in the ISCWSA MWD error models are representative of the magnitude of errors observed in practice. In addition to the ISCWSA defined error model terms, this research also includes an analysis of the sensor twist error term and the associated limits defined in the Fault Detection, Isolation, and Recovery (FDIR) error model. This study is comprised of 138 MWD runs that were selected based on the criteria that they were processed using FDIR with overlapping gyro surveying to ensure highly accurate and consistent estimated values. The error magnitudes and uncertainties estimated by FDIR were compiled and analyzed in comparison to the expected limits outlined in the error models. The results conclude that the limits defined in the ISCWSA error models are not always representative of what is observed in practice. For instance, in U.S. land the assumed magnitudes of several of the error sources are overly optimistic compared to the values observed in this study. This means that EOUs with which wells are planned may not be large enough in some scenarios which could cause the operator to assume unanticipated additional risk. The final portion of this analysis was undertaken to test the hypothesis that preventative measures such as additional non-magnetic spacing are generally being taken by operators and directional service providers to minimize additional injected error when survey corrections are not being implemented while drilling the well. This hypothesis was tested by dividing the 138 MWD runs into Historical (survey corrections were not utilized in real-time) and Real-Time (survey corrections were utilized in real-time) categories. The results indicate that there are no significant differences in the error estimates between the Historical and Real-Time categories. This result in combination with the determination that the majority of the error model error terms should be categorized as fat-tail distributed indicate that proper well spacing and economics calculated using separation factor alone are insufficient without the use of survey corrections in Real-Time.

Consistency Index-Based Sensor Fault Detection System for Nuclear Power Plant Emergency Situations Using an LSTM Network

A nuclear power plant (NPP) consists of an enormous number of components with complex interconnections. Various techniques to detect sensor errors have been developed to monitor the state of the sensors during normal NPP operation, but not for emergency situations. In an emergency situation with a reactor trip, all the plant parameters undergo drastic changes following the sudden decrease in core reactivity. In this paper, a machine learning model adopting a consistency index is suggested for sensor error detection during NPP emergency situations. The proposed consistency index refers to the soundness of the sensors based on their measurement accuracy. The application of consistency index labeling makes it possible to detect sensor error immediately and specify the particular sensor where the error occurred. From a compact nuclear simulator, selected plant parameters were extracted during typical emergency situations, and artificial sensor errors were injected into the raw data. The trained system successfully generated output that gave both sensor error states and error-free states.