Machine Maintenance Decision Model With Expert Feedback Learning
Find Innovative SolutionsGenerate Solutions
Solution Overview
Problem
Current health monitoring algorithms for aircraft engines are poorly calibrated due to lack of degradation data, leading to inefficient maintenance operations and increased risk of false alarms, as they rely on anomaly detection models based on normal cases rather than observed degradation signatures.
Innovation Solution
A decision aid system that uses a learning decision model to automatically compute operational diagnoses by correlating health indicators from anomaly detection modules with expert diagnoses, capable of iterative relearning to minimize diagnosis errors and improve anomaly detection accuracy.
Engineering Contradictions & Design Principles
Engineering Contradiction Analysis
1Reliability
If anomaly detection algorithms based only on normal cases are used, then the system can operate without degradation data, but the ability to identify types of degradation is poor
Solution Approach 1:
The system performs preliminary action by collecting and storing degradation data during engine operations before actual degradation events occur. Health indicators are continuously monitored and stored in a database, creating a prepared dataset that can be used for training anomaly detection algorithms when degradation events are eventually detected, thus improving detection accuracy without requiring wait for actual failures.
Solution Approach 2:
The system implements feedback by using expert diagnoses to validate and refine the anomaly detection algorithms. Expert opinions on degradation types are fed back into the system to improve the calibration of health monitoring algorithms, creating a continuous improvement loop that enhances degradation identification capability over time while maintaining operation with limited initial degradation data.
2Reliability
If experts verify each alert manually, then false alarms can be identified, but the time required for decision-making increases significantly
Solution Approach 1:
The system introduces an intermediary layer between anomaly detection and expert verification by implementing a decision support system that pre-processes alerts using multiple criteria (health indicators, operational context, historical data). This intermediary filters and prioritizes alerts before presenting them to experts, maintaining verification accuracy while significantly reducing the number of alerts requiring manual review and thus decreasing overall decision-making time.
Solution Approach 2:
The system enables self-service by allowing the anomaly detection system to automatically assess and prioritize its own alerts based on learned patterns from historical data and expert feedback. The system can autonomously handle routine alerts with high confidence, reducing the burden on experts to verify every single alert manually, thus improving efficiency while maintaining reliability for critical cases.
3Reliability
If engines are repaired before damage occurs, then safety is maintained, but degradation data for calibration is lost
Solution Approach 1:
The system performs preliminary action by continuously collecting and archiving health indicator data during normal engine operations and early degradation stages, before repairs are executed. This creates a preserved dataset of degradation trajectories that can be used for algorithm calibration, allowing the system to maintain both engine safety through timely repairs and retain valuable degradation information for improving detection accuracy.
Solution Approach 2:
The system applies discarding and recovering by intentionally preserving degradation data that would otherwise be lost when engines are repaired. The collected health indicators and operational data are stored in a dedicated database structure designed to retain degradation trajectories, effectively 'recovering' the information value from what would normally be discarded during routine maintenance operations.
Data Source
AI summary
A decision aid system, method and computer program product for the maintenance of a machine, including anomaly detection modules to determine health indicators on the basis of measurements of physical parameters of the machine, a calculator to compute an operating diagnosis on the basis of health indicators by applying a decision model capable of learning, and a man-machine interface to allow an expert to consult the health indicators and to declare a diagnosis. The calculator can compare an operational diagnosis computed based on a set of health indicators with at least one expert diagnosis declared after consultation of the set of health indicators, and can modify the decision model in the event of disagreement between the anomaly diagnosis computed and an expert diagnosis declared.