Composite Confidence Interval for Fault Diagnostics
Find Innovative SolutionsGenerate Solutions
Solution Overview
Problem
Existing health and usage management systems (HUMS) for complex systems like aircraft face challenges in accurately predicting and detecting component fault conditions due to inaccurate characterization of fault indicators, leading to decreased detection accuracy and increased false alarm rates.
Innovation Solution
A method and system that generate a composite confidence interval based on sampled subsets of sensor data from components with known operational status, using iterative sampling with replacement to determine upper and lower bounds, which helps in accurately representing the distribution of features among a population of similar components, thereby improving fault diagnosis accuracy.
Engineering Contradictions & Design Principles
Engineering Contradiction Analysis
1Measurement precision
If traditional fault threshold comparison methods are used, then the system can detect fault conditions, but the detection accuracy decreases and false alarm rates increase due to inaccurate characterization of fault indicators
Solution Approach 1:
The system performs preliminary characterization of fault indicators by collecting sensor data from components with known operational status and generating composite confidence intervals before actual fault detection occurs. This preparatory statistical modeling enables more accurate fault detection by establishing reliable baseline expectations of normal operation, thereby reducing both detection errors and false alarms during actual monitoring
Solution Approach 2:
The patent introduces composite confidence intervals as an intermediary statistical construct between raw sensor data and fault detection decisions. These confidence intervals serve as a mediator that characterizes the distribution of fault indicators and provides a probabilistic framework for comparing against actual measurements, improving detection accuracy while reducing false alarms through statistically rigorous thresholds
2Measurement precision
If composite confidence intervals are generated through iterative sampling, then fault diagnosis accuracy improves, but the computational time and processing resources increase
Solution Approach 1:
The system performs the computationally intensive iterative sampling and composite confidence interval generation in advance, during periods when components are known to be operating normally. By completing these calculations beforehand and storing the results, the system avoids repeating expensive computations during actual fault detection, thus improving diagnosis accuracy while minimizing real-time processing time losses
Solution Approach 2:
The system dynamically adapts its computational effort by performing extensive iterative sampling only when generating baseline confidence intervals from known good data, rather than continuously during operational monitoring. This dynamic approach concentrates computational resources where they provide maximum value (characterization phase) while minimizing time loss during routine fault detection operations
Data Source
Figure 1
Figure 2
Figure 3
AI summary
In one example, a method includes measuring sensor data of at least one physical component having a known operational status. The method further includes generating a plurality of data points from the measured sensor data, each of the plurality of data points representing a measured occurrence of a feature of the measured sensor data. The method further includes iteratively sampling with replacement the data points to generate a plurality of subsets of the data points, and determining, within each of the plurality of subsets, a confidence interval having an upper bound and a lower bound to generate a plurality of confidence intervals having respective upper bounds and lower bounds. The method further includes generating a composite confidence interval having a composite upper bound based on a first central tendency of the respective upper bounds and a composite lower bound based on a second central tendency of the respective lower bounds.