Rocket Engine Fault Detection Using Hierarchical Confidence Levels
Find Innovative SolutionsGenerate Solutions
Solution Overview
Problem
Current fault detection systems for complex systems, such as rocket engines, rely on majority logic approaches that are inadequate for handling diverse and severe malfunctions, often leading to untimely shutdowns and inadequate responses due to lack of confidence and severity consideration.
Innovation Solution
A fault detection and location system that uses sensors, validation modules, fault detection modules, data fusion modules, and decision-making modules to generate confidence-level messages, allowing for precise diagnosis and adaptive action based on severity and reliability, incorporating fuzzy logic for nuanced decision-making.
Engineering Contradictions & Design Principles
Engineering Contradiction Analysis
1Device complexity
If majority logic approach is used for fault detection, then system complexity is reduced, but measurement precision and reliability of fault diagnosis deteriorate
Solution Approach 1:
The fault detection system is segmented into multiple hierarchical levels: sensor level (first order confidence), component level (second order confidence), and system level (third order confidence). Each level processes fault information independently and contributes to the overall diagnosis, allowing precise fault localization without requiring complex global analysis.
Solution Approach 2:
The patent introduces a new dimension for fault analysis by incorporating confidence levels (first order, second order, third order) that represent different hierarchical levels of the system. This multi-dimensional approach allows the system to evaluate faults not just by presence/absence but by their reliability and hierarchical context, significantly improving diagnostic precision.
2Loss of time
If majority logic approach is used for fault detection, then response time is reduced, but reliability of action decision deteriorates due to lack of confidence and severity consideration
Solution Approach 1:
The system performs preliminary classification of faults by assigning confidence levels (first order from sensors, second order from components, third order from system level) before making final decisions. This preliminary assessment of fault reliability allows the system to prepare appropriate responses in advance, reducing response time while maintaining high decision reliability through structured confidence evaluation.
Solution Approach 2:
The patent changes the parameter representation from binary fault detection to multi-level confidence assessment. By transforming fault information into graded confidence levels (first order, second order, third order) and combining them with severity parameters, the system achieves both rapid response and reliable decision-making through enriched parameter analysis.
3Reliability
If confidence levels and severity considerations are added to fault detection, then action decision reliability is improved, but device complexity increases
Solution Approach 1:
The complex task of reliable fault diagnosis is segmented into three manageable confidence levels: first order (sensor validation), second order (component fault detection), and third order (system-level data fusion). Each segment handles a specific aspect of confidence assessment, making the overall complex system more manageable and implementable while achieving high decision reliability.
Solution Approach 2:
The patent adds the dimension of hierarchical confidence levels (first order, second order, third order) to the fault detection process. This dimensional expansion allows the system to capture multiple aspects of fault reliability without requiring exponentially complex processing, as each level builds upon the previous one in a structured hierarchy.
Data Source
Figure 1~8
Figure 4
Figure 5A~6B
AI summary
The invention relates to a system for detecting and locating breakdowns that can occur in a complex system, comprising: means (CAP, MODVAL,MODDP, MODFD,) for detecting and locating a breakdown affecting at least one subsystem of the complex system; and means (MODPD) for making a decision regarding possible action to be taken in the event of such a detection. Said system is of use in monitoring a rocket engine.