Virtual Flow Meter Sensor Fault Isolation Using PCA
Find Innovative SolutionsGenerate Solutions
Solution Overview
Problem
Virtual flow meters in resource production contexts, such as oil and gas, face challenges in maintaining accuracy due to model errors and sensor failures, particularly in subsea locations where sensor biases, drifts, and failures compromise measurement quality.
Innovation Solution
A data-driven multivariate statistical method employing Principal Components Analysis, Weighted Squared Prediction Error, and Partial Decomposition Contribution Plots is used for fault detection and isolation in virtual flow metering, enhancing sensor measurement robustness and calibration.
Engineering Contradictions & Design Principles
Engineering Contradiction Analysis
1Measurement precision
If virtual flow meters use complex data-fusion algorithms and sensor systems for estimating flow rates, then measurement accuracy is improved, but system complexity and susceptibility to sensor failures increase
Solution Approach 1:
The patent segments the sensor system into multiple independent sensors (pressure sensors, temperature sensors) that can be individually monitored and diagnosed. This segmentation allows the system to identify and isolate faulty sensors without complete system failure, resolving the contradiction by maintaining measurement accuracy through redundant independent components while managing complexity through modular architecture
Solution Approach 2:
The patent implements preliminary fault detection and isolation mechanisms that proactively identify sensor failures before they compromise measurement accuracy. By establishing monitoring algorithms and fault detection thresholds in advance, the system maintains accurate flow measurements while preventing the propagation of errors from complex interconnected sensors
2Duration of action of stationary object
If virtual flow meters operate over extended periods in subsea environments, then production continuity is improved, but sensor measurement quality deteriorates due to bias, drift, and failure
Solution Approach 1:
The patent establishes a training phase during which the system learns normal sensor behavior patterns and establishes baseline characteristics. This preliminary action enables the system to detect deviations indicating sensor drift or bias over time, allowing for maintained measurement quality throughout the extended production site lifespan by identifying and compensating for sensor degradation
Solution Approach 2:
The patent implements continuous monitoring and feedback mechanisms that track sensor performance over the production site lifespan. By comparing current measurements against established baselines and detecting trends in sensor drift or bias, the system maintains measurement precision over extended periods through adaptive compensation and fault isolation
Data Source
Figure 1
Figure 2
Figure 3
AI summary
Technical effects of the invention include use of a data-driven multivariate statistical method for the detection and isolation of sensor faults applied in a virtual flow metering context. In one implementation, the data-driven multivariate statistical method employs principal components analysis, weighted squared prediction error, and partial decomposition contribution plots.