304 / 2025-06-19 19:52:20

An Anomaly Analysis Method for Rocket Telemetry Data Based on SincNet-LSTM

LSTM, rocket telemetry data, anomaly detection, trend forecasting

Combining the respective advantages of the SincNet neural network and the Long Short-Term Memory (LSTM) network, an anomaly analysis method for rocket telemetry data based on SincNet-LSTM is proposed. SincNet is utilized to precisely identify and extract feature vectors from the time-series structure of rocket telemetry data. The feature vectors output by the SincNet layer, along with the corresponding target vectors, are used as inputs. After training the LSTM network model, prediction results are obtained. The average error between the prediction results and the original data is calculated. Then, based on a non-parametric threshold calculation method, a threshold is determined to judge whether the telemetry data is anomalous, ultimately achieving rocket anomaly detection. The feasibility and effectiveness of the SincNet-LSTM analysis method are verified using telemetry data from a certain type of rocket.