Aircraft Sensor Fault Detection Using Residual Pattern Recognition
Find Innovative SolutionsGenerate Solutions
Solution Overview
Problem
Aircraft pitot tubes are prone to measurement failures due to icing blockage, heavy water ingestion, and volcanic ash blockage, leading to common mode pneumatic events that corrupt airspeed displays and potentially cause aircraft deviation from optimal flight paths, necessitating an improved sensor fault detection and identification technology.
Innovation Solution
A method and system using residual failure pattern recognition with statistical filters, such as extended Kalman filters, to detect and identify sensor failures in aircraft sensors like pitot tubes, static ports, and accelerometers, by removing known corruption effects and generating alert signals and synthetic data signals to ensure accurate flight control.
Engineering Contradictions & Design Principles
Engineering Contradiction Analysis
1Reliability
If traditional sensor monitoring is used, then the system is simple, but sensor failures cannot be detected and identified in time
Solution Approach 1:
The patent introduces residual signals as an intermediary element that mediates between the sensor measurements and the fault detection system. These residual signals are generated by comparing actual sensor readings with expected values from a mathematical model, serving as a bridge that translates sensor data into fault indicators without requiring direct modification of the sensors themselves
Solution Approach 2:
The patent replaces traditional mechanical or electronic fault detection mechanisms with a signal processing and mathematical modeling approach. Instead of using physical sensors to detect sensor failures, the system uses residual analysis and pattern recognition algorithms to identify faults, substituting mechanical/electronic detection with computational methods
2Reliability
If sensor failures are not detected, then the system operates normally, but corrupted airspeed data causes aircraft deviation from optimal flight path
Solution Approach 1:
The patent performs preliminary fault detection by continuously monitoring residual signals before corrupted sensor data can significantly impact flight operations. The system proactively identifies sensor failures through residual pattern analysis, allowing corrective action to be taken before the faulty data causes aircraft deviation or safety issues
Solution Approach 2:
The patent implements a feedback mechanism where residual signals are continuously generated from sensor measurements, analyzed for fault patterns, and used to trigger alerts or switch to backup sensors. This closed-loop feedback system ensures that flight path accuracy is maintained by rapidly responding to sensor failures and preventing corrupted data from causing aircraft deviation
3Reliability
If multiple sensors are monitored individually, then comprehensive coverage is achieved, but the complexity of fault identification increases
Solution Approach 1:
The patent merges the monitoring of multiple sensors into a unified fault detection framework by combining their residual signals into a common analysis structure. Instead of treating each sensor independently, the system integrates their residuals and applies pattern recognition to identify faults across the sensor array, reducing the complexity of fault identification while maintaining comprehensive coverage
Solution Approach 2:
The patent creates a universal fault detection mechanism that handles multiple sensor types (pitot tubes, static ports, accelerometers, gyroscopes) through a single residual analysis system. The pattern recognition approach is designed to be multi-functional, capable of identifying faults in any monitored sensor by analyzing characteristic residual patterns, eliminating the need for separate detection systems for each sensor type
Data Source
Figure 1A
Figure 1B
Figure 2A~2B
AI summary
Systems, methods, and apparatus for sensor fault detection and identification using residual failure pattern recognition are disclosed. In one or more embodiments, a method for sensor fault detection and identification for a vehicle comprises sensing, with sensors (120, 130, 140, 150) located on the vehicle (110), data. The method further comprises performing majority voting on the data (325a ... 325d) for each of the types of data to generate a single voted value (350a ... 350d) for each of the types of data. Also, the method comprises generating, for each of the types of data, estimated values (420a ... 420d) by using some of the voted values. In addition, the method comprises generating residuals (r1 ... r4) by comparing the estimated values to the voted values. Further, the method comprises analyzing a pattern of the residuals to determine which of the types of the data is erroneous to detect and identify a fault experienced by at least one of the sensors on the vehicle.