Deep learning identification method for helicopter flight attitude imbalance data

Abstract

The invention discloses a deep learning identification method for helicopter flight attitude imbalance data, which comprises the following steps: collecting helicopter flight attitude data and carrying out normalization processing, carrying out sample division on the processed flight attitude data, and dividing the divided samples into a training set and a test set according to a category imbalance rate; and constructing a convolutional neural network to obtain a predicted value of each category of data, wherein the convolutional neural network comprises a convolutional layer, a pooling layer and a full connection layer; subjecting the input matrix to feature extraction through a convolution layer, inputting the extracted features to a pooling layer for feature dimension reduction and further performing feature extraction after being subjected to nonlinear activation, and inputting the obtained features into a full connection layer to obtain a predicted value of each category of data; and constructing focusing loss, taking the focusing loss as a loss function of the convolutional neural network, and adjusting the weight of the sample through a weight factor and a focusing factor in the focusing loss. Therefore, the network can extract valuable information under the condition that the samples are unbalanced, and the recognition precision of the network under the condition that the samples are unbalanced is improved.

Description

technical field [0001] The disclosure belongs to the field of helicopter flight attitude recognition, in particular to a deep learning recognition method for unbalanced helicopter flight attitude data. Background technique [0002] Helicopters, as the main aviation aircraft flying at low altitude, play a pivotal role in both the national military security field and the people's livelihood industry field. Due to the flexible and maneuverable characteristics of the helicopter, its flight attitudes are diverse. In different flight attitudes, the loads carried by the moving parts and life parts of the helicopter are complex and changeable, which is likely to cause differences in damage, resulting in frequent failures of the helicopter. The correct identification of the attitude of the helicopter aircraft can establish a one-to-one correspondence between the various components of the helicopter and its measured load spectrum, which provides an important basis for the life predict...

AI Technical Summary

Problems solved by technology

However, due to the difficulty in obtaining the actual flight data of the helicopter and the characteristics of the flight mode, it is often difficult to obtain enough samples for some actions, resulting in data imbalance between various flight attitudes. Therefore, in the training of the intelligent diagnostic model, these A small number of sample categories will naturally fall into the predicament of underfitting, making it impossible to accurately identify the flight state of the helicopter

Images