Aircraft Malfunction Probability Modeling with Causal Measurement Filtering
Find Innovative SolutionsGenerate Solutions
Solution Overview
Problem
Existing learning systems struggle to establish reliable correlations for rare aircraft malfunctions due to insufficient data, leading to ineffective database reduction or distorted statistics when attempting to predict performance defects.
Innovation Solution
A method involving data acquisition, grouping by root cause, causal time windowing, measurement weighting, Boolean value assignment, and subgrouping to refine input data for accurate probability determination, using a computing system to process and analyze sensor measurements.
Engineering Contradictions & Design Principles
Engineering Contradiction Analysis
1Reliability
If the database size is reduced to have 50% actual incidents, then the learning system can be trained with balanced data, but the database size becomes too small to allow reliable training
Solution Approach 1:
The patent segments the database into multiple groups based on malfunction types, aircraft types, and operational conditions. Each segment is analyzed separately with its own learning model, allowing sufficient data volume within each segment while maintaining overall database integrity and training reliability.
Solution Approach 2:
The patent introduces additional dimensions for data organization including temporal dimensions (time windows), hierarchical dimensions (malfunction categories), and contextual dimensions (operational conditions). This multi-dimensional segmentation enables adequate sample sizes in each dimension while preserving the rare incident data.
2Quantity of substance
If cases are artificially cloned to increase the number of cases, then the database size increases for training, but the statistics of malfunction occurrence are distorted
Solution Approach 1:
The patent introduces synthetic intermediate cases generated through physics-based simulations and digital twins as mediators between actual rare incidents and training requirements. These synthetic cases preserve statistical accuracy while providing sufficient training data volume by modeling plausible malfunction scenarios without duplicating actual incidents.
3Loss of information
If all measurements are retained for analysis, then complete data is available for correlation, but the processing complexity and computational burden increase
Solution Approach 1:
The patent extracts and selects only the most relevant measurements and features that have proven correlations with malfunction occurrences. Feature selection techniques identify key parameters from the full measurement set, removing redundant data while preserving critical information for accurate prediction.
Solution Approach 2:
The patent performs preliminary data processing including filtering, aggregation, and feature engineering before main analysis. Time-window-based preprocessing consolidates raw measurements into meaningful indicators, reducing data volume while maintaining predictive value for malfunction detection.
Data Source
AI summary
A method for determining a probability of occurrence of a malfunction creating a performance defect in an aircraft makes it possible to process a set of input data so as to retain only measurements relevant with regard to determining the probability of occurrence of the malfunction following a previous event. In other words, the method makes it possible to refine a set of input data so as to establish relevant causal links between records of measurements and a malfunction, so as to determine a probability of occurrence of a malfunction based on the occurrence of one or more previous events.