Research on the Stability of a Machine Learning Model for Processing a Signal of the Earth’s Orientation Device

• Published in: Mathematical models and computer simulations 2023-12, Vol.15 (6), p.1052-1058
• Main Authors: Sinyutin, S. A., Sinyutin, E. S., Yartsev, A. V.
• Format: Article
• Language: English
• Subjects: Algorithms; Classifiers; Decision trees; Earth; Machine learning; Mathematical Modeling and Industrial Mathematics; Mathematics; Mathematics and Statistics; Noise generation; Orientation; Signal processing; Signal to noise ratio; Simulation and Modeling; Stability
• ISSN: 2070-0482; 2070-0490
• DOI: 10.1134/S2070048223060169

This article considers the solution of the problem of recognizing the artifacts of the Sun and the Moon in the signal of the photodetector (PD) of the Earth’s orientation device. The previously developed machine learning model, trained on clean signals (real and generated), must be tested for robustness (stability) to real signals that can be noisy and the model should be corrected if necessary. The stability of the machine learning model to noisy signals is studied, and a decrease in the quality of the classifier of normal and artifact signals of the PD of the Earth’s orientation device is revealed. To correct the model, the method of augmenting the initial data using artificially noisy signals is used, the types of noise characteristic of electronics (thermal-, shot-, and flicker-noise) are selected, and the levels of the signal-to-noise ratio (SNR) for devices of this class are selected. Methods for generating noise signals are selected and algorithms for the correct mixing of clean signals with noise are developed. The shortcomings of the existing linear model, which appear only when using different levels of the SNR, are revealed. The linear model’s errors turn out to be significant for the application area. A new classifier model based on decision trees is selected. The new model’s stability is also tested, and the need to expand the range of the SNR for the training sample is shown. The applied method of classifying artifacts and increasing the stability of the model can be used in PDs if it is necessary to provide built-in signal processing in its computing core of the device.