Machine learning sample generation method for aircraft thrust fault on-line identification

Abstract

The invention relates to a machine learning sample generation method for aircraft thrust fault on-line identification, and the method is suitable for the field of typical power system fault on-line identification in the aircraft flight process. Data fusion generation is carried out on flight motion information (such as flight position, speed, acceleration, attitude angle and angular speed) of a control system, and corresponding data is intercepted as a machine learning training and testing sample according to the design method provided by the invention. Factors such as mass center motion, disturbance center motion, structural disturbance, aerodynamic force and moment of force of the aircraft are considered, deviation combination circulation is introduced into the simulation model to generate data, the data is more real and credible, and improvement of actual fault identification precision is facilitated. According to the method, the fault mode is refined, related data with relatively fine granularity of the fault mode is generated, and the identification precision is improved.

Description

technical field [0001] The invention relates to a machine learning sample generation method for online identification of aircraft thrust faults, which is suitable for the field of online identification of faults during the flight of launch vehicles. Background technique [0002] Due to limited real data, aircraft fault identification technology based on machine learning mostly relies on aircraft simulation models to generate a large number of machine learning samples. If the model is inaccurate or has a large gap with the real model, it will greatly affect the accuracy of actual flight identification. Existing machine learning fault identification can achieve high accuracy in simulation tests, but the effect is often not good in practical applications, mostly because of the large gap between machine learning sample data and real data. Moreover, the current machine learning sample data is large in scale, takes up large computing resources, affects computing efficiency, and is...

AI Technical Summary

Problems solved by technology

[0002]Due to limited real data, aircraft fault identification technology based on machine learning mostly relies on aircraft simulation models to generate a large number of machine learning samples. If the model is inaccurate or far from the real model , it will greatly affect the actual flight identification accuracy

Existing machine learning fault identification can achieve high accuracy in simulation tests, but the effect is often not good in practical applications, mostly due to the large gap between machine learning sample data and real data

Moreover, the current machine learning sample data is large in scale, takes up large computing resources, affects computing efficiency, and is not suitable for computing hardware with limited computing power

Images