Real-time automatic detection of signals triggered by space dust's impact based on deep learning

Accurate and rapid detection of dust impact events on spacecraft can help us better understand the dust distribution in the space and reduce the damage to spacecraft due to dust impacts. Although the existing methods of manual identification or machine identification of dust impact events based on the waveform characteristics of potential difference signals caused by dust impacts have high accuracy, their efficiency is low, and high-precision and automated methods are urgently needed to identify the massive potential difference signals collected by spacecraft. The deep learning model has strong ability in signal classification and recognition. In this paper, the problem of potential difference signals caused by dust impacts detection is modeled as a signal classification problem, and a convolutional neural network model is constructed, which can automatically extract signal features and classify signals according to the features. At the same time, in order to train the model and test the prediction accuracy of the model, a data set composed of potential difference signals caused by dust impacts and potential difference signals caused by other events was constructed. The accuracy rate of the model on training set is 99.46% and on the test set is 98. 68%, the recall rate is 99.44%, the precision rate is 97.95%, and the threat score is 97.41%, High-precision and automatic dust collision events detection is realized.