Air Data Estimation Using Internal Avionics and Neural Networks
Find Innovative SolutionsGenerate Solutions
Solution Overview
Problem
Conventional air data systems on aircraft rely on external sensors that are vulnerable to damage, create aerodynamic drag, and are challenging to integrate on high-performance military aircraft, necessitating an alternative method for accurate air data parameter estimation without external probes.
Innovation Solution
A computer-implemented method and apparatus that uses internal avionics data and empirical data from wind tunnel testing and computational fluid dynamics to simulate air data parameters, including turbulence, and trains a neural network for real-time air data estimation, eliminating the need for external sensors.
Engineering Contradictions & Design Principles
Engineering Contradiction Analysis
1Measurement precision
If external sensors are used for air data measurement, then measurement precision is improved, but device complexity and vulnerability to damage increase
Solution Approach 1:
The patent extracts the air data measurement function from external sensors and relocates it to internal avionics systems. By using existing internal sensors (accelerometers, gyroscopes, flight control data) to compute air data parameters through simulation and machine learning, the system eliminates the need for vulnerable external air data probes while maintaining measurement capability.
Solution Approach 2:
The patent creates a virtual copy of external sensor functionality through computational simulation. Instead of physically measuring air data with external probes, the system simulates air data parameter values by processing internal avionics data through trained neural networks, producing equivalent measurement results without physical exposure to external environmental hazards.
2Measurement precision
If external sensors are installed on aircraft, then air data parameters can be measured, but aerodynamic drag increases
Solution Approach 1:
The measurement function is extracted from external aerodynamic-exposed sensors and relocated to internal avionics. The system uses internal sensors that do not protrude into the airflow, eliminating the aerodynamic drag caused by external probes while maintaining air data measurement capability through computational methods.
3Measurement precision
If external sensors are used for air data measurement, then measurement accuracy is improved, but reliability decreases due to vulnerability to damage
Solution Approach 1:
The measurement capability is extracted from vulnerable external sensors and relocated to protected internal avionics systems. By computing air data parameters from internal sensor data through simulation and machine learning models, the system eliminates exposure to environmental damage while preserving measurement accuracy.
Solution Approach 2:
The patent employs machine learning models trained on extensive simulation data to compensate for potential data gaps or anomalies. The trained neural networks provide robust predictions that cushion against uncertainties, ensuring reliable air data estimation even when input data quality varies, thereby enhancing system reliability.
4Productivity
If simulation data is used for training, then productivity is improved, but manufacturing precision requirements increase
Solution Approach 1:
The patent performs preliminary actions by generating extensive simulation training data before deploying the machine learning model. Comprehensive simulation campaigns pre-compute air data parameter relationships across the full flight envelope, creating a robust training dataset that enables the neural network to achieve high accuracy without requiring expensive physical flight tests.
Solution Approach 2:
The system creates computational copies of physical flight test data through high-fidelity simulation. By replicating flight conditions and sensor responses in virtual environments, the patent generates training data that mirrors real-world scenarios, achieving both productivity gains and the precision needed for accurate model training.
Data Source
Figure 1
Figure 2
Figure 3
AI summary
A computer-implemented method of computing training data for training an air data estimator to predict values of air data parameters of an aircraft comprises: selecting ground truth values of the air data parameters according to a flight envelope of the aircraft. The method involves simulating corresponding values from avionics in the aircraft by using stored empirical data about engines of the aircraft and empirical data about the aircraft obtained from wind tunnel testing, and rules of computational fluid dynamics.