Aviation dynamotor fault diagnosis method based on FWA-RNN model

Abstract

The invention relates to an aviation dynamotor fault diagnosis method based on a FWA-RNN model, and belongs to the dynamotor state monitoring and fault diagnosis field; the method mainly comprises thefollowing steps: 1, analyzing fault modes and selecting test signals; 2, determining an aviation dynamotor to-be-tested signal, using sensors to gather mass data under various work states, and sending the data to a computer for data storage; 3, making FFT transformation for obtained mass data, carrying out normalization processing, and splitting obtained data into training samples and test samples; 4, combining test sample data, using FWA to optimize all initial parameters of a RNN, building a nerve network model, and inputting the training samples so as to make fault analysis; 5, detecting the model accuracy. The method has excellent data adaptive ability and robustness, and can effectively improve the aviation dynamotor fault classification detection efficiency and accuracy.

Description

technical field [0001] The invention relates to an aviation generator fault diagnosis method based on an FWA-RNN (Fireworks Algorithm-Recursive Neural Network) model, belonging to the field of generator state monitoring and fault diagnosis. Background technique [0002] The generator is the core of the power system. Once a failure occurs, it will not only threaten the stable operation of the power system, but also damage the generator equipment. With the continuous increase of stand-alone capacity, the operational reliability of generators is particularly important and prominent. Especially in the aviation power system, the normal operation of the brushless alternator has a decisive impact on the vitality of the aircraft. Therefore, it is of great significance to the development of the aviation industry to study and analyze the fault diagnosis of aero-generator equipment, to propose a more effective fault diagnosis system, and to better ensure the safety of aircraft operati...

AI Technical Summary

Problems solved by technology

At present, the commonly used fault diagnosis methods mainly adopt manual analysis or signal processing methods, and these methods have some shortcomings, such as low analysis efficiency, low degree of automation, etc.

[0004] Recursive neural network is a kind of neural network with fixed weights and thresholds. Because of the stability of its internal structure, it is easier to realize. It has the characteristics of simple principle, fast running speed, and strong local search ability. However, the initial parameter setting Improper use may cause it to converge slowly, or fall into a local minimum

Images