Multimodel Fault Detection Using Constrained Kalman Filtering
Find Innovative SolutionsGenerate Solutions
Solution Overview
Problem
Existing fault detection methods in industrial systems often lead to false failures due to operational changes and lack of robustness in handling modeling errors and measurement noise, particularly in petrochemical industries where reliability and safety are critical.
Innovation Solution
A multimodel fault detection and diagnosis system utilizing a constrained Kalman filter (CKF) supervisor that generates a bank of submodels for both healthy and faulty scenarios, allowing for the use of analytical, data-driven, or combined models, and effectively handles modeling errors and noise by projecting estimates onto constraint spaces.
Engineering Contradictions & Design Principles
Engineering Contradiction Analysis
1Device complexity
If a single reference model is used for fault detection, then the detection method is simple, but false failures are detected due to operational changes and natural moves between operating regions
Solution Approach 1:
The system segments the operational space into multiple operating regions, each with its own reference model. This segmentation allows the system to adapt to natural transitions between operating conditions without triggering false alarms, while maintaining simple detection logic within each region.
Solution Approach 2:
The system dynamically switches between different reference models based on the current operating region. This dynamic adaptation enables the fault detection system to follow natural operational changes while maintaining reliability, avoiding the static limitations of a single reference model.
2Reliability
If multiple models are used to cover different operating regimes, then fault detection reliability is improved, but device complexity increases
Solution Approach 1:
The operational space is segmented into distinct regions, each governed by a simple reference model. This segmentation reduces the complexity of individual models while maintaining overall reliability through comprehensive coverage of all operating conditions.
Solution Approach 2:
The system employs a universal framework that can accommodate multiple reference models within a single fault detection architecture. This multi-functional approach allows different models to work together harmoniously, improving reliability without proportionally increasing overall system complexity.
3Reliability
If operational data is processed through constrained Kalman filter with model bank, then measurement noise and modeling errors are handled effectively, but computational complexity increases
Solution Approach 1:
The system applies partial filtering through the constrained Kalman filter, focusing computational resources on the most critical aspects of noise and error handling. This selective approach provides sufficient robustness without the full computational burden of exhaustive filtering.
Solution Approach 2:
The constrained Kalman filter acts as an intermediary between the multiple reference models and the fault detection logic. This mediator handles the computational complexity of noise and error management, allowing the individual models to remain relatively simple while achieving robust overall performance.
Data Source
AI summary
In a method for failure, detection, operational data from a plurality of components in a system are received. A bank of submodels is created based on the operational data. The bank of submodels corresponds to a normal mode and one or more faulty modes of the system and is valid in different operating regimes of the system. Each of the banks of submodels has a respective weight (validity) and a respective suboutput. An output of the system is a weighted sum of the suboutputs of the submodels. The operational data is therefore processed to generate a validity profile through a constrained Kalman Filter (KCF) based multimodel fault detection and diagnosis (FDD). Subsequently, the validity profile is output. The validity profile is indicative of an operation state of the system at a given time, and the operation state includes a normal state and a fault state.