Filtering of noise interference in piezo sensor signals by means of neural networks in on-board space equipment

The article discusses the problem of filtering noise interference in the signals of piezoelectric sensors used in on-board space equipment. Presents the results of experimental studies on the use of three types of neural networks: direct propagation neural network (FFNN), autoencoder (AE) and convolutional neural networks (1D CNN). The main results showed that convolutional neural networks demonstrate the best filtering performance of high-frequency interference while maintaining the basic characteristics of the signal. The conducted research allowed us to determine the optimal learning parameters for each architecture of the neural network. The practical significance of the work lies in the possibility of using the developed methods to increase the reliability of spacecraft monitoring systems and protect against micrometeoroid hazards. The results obtained can be used in the development of the latest spacecraft surface control systems.

Authors: D. P. Grigoriev, K. E. Voronov

Direction: Informatics, Computer Technologies And Control

Keywords: piezoelectric sensors, spacecraft, microcontroller, Python, Tensorflow, Keras, neural networks, filtering, AM signals, microparticles

---

View full article