Neural Air Data Estimation for Pitot-Independent Aircraft Airspeed
Find Innovative SolutionsGenerate Solutions
Solution Overview
Problem
Aircraft airspeed estimation is critical but prone to errors due to failures in traditional air data sensors, such as pitot tubes, which can be blocked by environmental factors, leading to inaccurate readings and potential catastrophic failures, especially in Fly-by-wire systems where redundancy may not suffice to prevent common cause faults.
Innovation Solution
A software-based neural network trained on aircraft model simulations and real flight data to estimate air data independently of traditional sensors, using existing aircraft systems without adding weight, cost, or complexity, by processing inputs from diverse sensors like GPS, IRU, and FADEC.
Engineering Contradictions & Design Principles
Engineering Contradiction Analysis
1Reliability
If traditional air data sensors (pitot tubes) are used for airspeed measurement, then the system structure is simple and cost-effective, but the reliability deteriorates due to susceptibility to blockage by ice crystals and environmental factors
Solution Approach 1:
The patent replaces the mechanical pitot tube sensor system with a software-based neural network system that processes data from existing aircraft sensors (GPS, IRU, FADEC, accelerometers). This substitution eliminates the physical components susceptible to blockage while maintaining airspeed measurement capability through computational estimation, thereby improving reliability without adding significant system complexity.
Solution Approach 2:
The neural network acts as an intermediary between the existing aircraft sensor systems and the airspeed parameter requirement. Instead of directly measuring airspeed through pitot tubes, the system uses the neural network to process intermediate data from multiple sensor sources (position, attitude, acceleration, engine parameters) and generate accurate airspeed estimates, thereby bypassing the reliability issues of traditional sensors.
2Reliability
If redundant air data sensors are installed to improve reliability, then the reliability improves through fault tolerance, but the device complexity and cost increase
Solution Approach 1:
The patent makes the existing aircraft sensor systems serve multiple functions. The GPS, IRU, FADEC, and accelerometer systems, originally designed for other primary functions (navigation, attitude reference, engine control, structural monitoring), are additionally utilized for airspeed estimation through the neural network. This multi-functionality provides redundancy and fault tolerance without installing dedicated redundant air data sensors, thereby improving reliability while avoiding increased system complexity.
3Reliability
If heating devices are added to prevent pitot tube blockage, then the reliability improves by preventing ice crystal accumulation, but the energy consumption and device complexity increase
Solution Approach 1:
The patent replaces the heating device mechanical/thermal protection system with a software-based alternative. Instead of actively preventing ice crystal accumulation through energy-consuming heating elements, the system uses the neural network to estimate airspeed from other sensor data, thereby eliminating the need for heating devices and their associated energy consumption while maintaining measurement reliability.
4Reliability
If a software-based neural network is implemented for air data estimation, then the reliability and independence from traditional sensors improve, but the device complexity increases due to additional processing systems
Solution Approach 1:
The neural network is implemented within existing aircraft flight control computers or avionics systems that already perform other computational functions. By utilizing existing processing hardware and integrating the airspeed estimation algorithm into existing software architectures, the patent avoids adding dedicated new hardware systems, thereby limiting the increase in device complexity while achieving improved reliability through independent software-based estimation.
Data Source
AI summary
Aircraft air data is estimated using a neural network trained to be independent of any signals from air data sensors whose values are based on air flow pressure measurements.