Abnormal Situation Prevention Algorithm for Plugged Line Diagnostics
Find Innovative SolutionsGenerate Solutions
Solution Overview
Problem
Existing abnormal situation prevention algorithms in process plants face challenges such as regression models becoming invalid when operating conditions change and the learning phase being too long relative to the time required for abnormal situations to develop, particularly in detecting plugged impulse lines in differential pressure flow measuring devices.
Innovation Solution
A system and method that calculate statistical signatures of process variables over sample windows, generating a function to model these signatures, and executing either a learning or monitoring function based on new data points, allowing for real-time prediction and detection of abnormal conditions by comparing predicted and actual values, with a mechanism to remove points from the model to maintain accuracy.
Engineering Contradictions & Design Principles
Engineering Contradiction Analysis
1Measurement precision
If traditional regression analysis is used to model monitored variable as a function of load variable during learning phase, then the model can predict values of monitored variable, but the learning phase becomes too long relative to the time required for abnormal situations to develop
Solution Approach 1:
The patent changes the parameters used in regression analysis from using raw process variables to using statistical signatures (mean, standard deviation, skewness, kurtosis) calculated over sliding windows. This transformation of parameters enables the model to capture abnormal conditions more quickly while maintaining prediction accuracy, reducing the learning phase duration significantly.
Solution Approach 2:
The patent implements a dynamic learning phase that automatically transitions to monitoring mode when sufficient data is collected. The system dynamically adjusts the learning period based on process stability criteria, allowing it to adapt to changing operating conditions without requiring a fixed, lengthy learning phase. This dynamic approach reduces the time to detect abnormal situations while maintaining model accuracy.
2Reliability
If regression model is trained during learning phase for abnormal situation detection, then the model can detect abnormal conditions, but the model becomes invalid when operating conditions change during monitoring phase
Solution Approach 1:
The system dynamically monitors process stability using statistical tests on the load variable and automatically re-enters learning mode when operating conditions change beyond predefined thresholds. This dynamic adaptation ensures the model remains valid under changing operating conditions while maintaining reliable abnormal situation detection. The system continuously adjusts between learning and monitoring modes based on real-time process behavior.
Solution Approach 2:
The patent implements feedback mechanisms where the system continuously evaluates the validity of the regression model by monitoring the residuals and process stability. When the model performance degrades or operating conditions change significantly, the system triggers a re-learning phase to update the model parameters. This feedback loop ensures the model adapts to new operating conditions while maintaining detection reliability.
3Adaptability or versatility
If statistical signatures are calculated over sliding windows with continuous model updating, then the system adapts to changing operating conditions, but the computational complexity increases
Solution Approach 1:
The patent segments the complex model updating task into distinct phases (learning phase and monitoring mode) with clear transition criteria. During monitoring mode, the system performs simpler validity checks rather than full model retraining. This segmentation reduces computational complexity while maintaining adaptability to operating condition changes through efficient phase transitions.
Solution Approach 2:
The system performs partial model updates by only recalculating statistical signatures and regression parameters when stability criteria are violated, rather than continuously updating the model at every data point. This partial action approach reduces computational burden while maintaining the system's ability to adapt to significant operating condition changes.
Data Source
AI summary
Systems and methods are provided for detecting abnormal conditions and preventing abnormal situations from occurring in controlled processes. Statistical signatures of a monitored variable are modeled as a function of the statistical signatures of a load variable. The statistical signatures of the monitored variable may be modeled according to an extensible regression model or a simplified load following algorithm. The systems and methods may be advantageously applied to detect plugged impulse lines in a differential pressure flow measuring device.