A Method for Analyzing Time Series Data Characteristics of Turbofan Engines Based on Adaptive Entropy Value

Health monitoring and remaining life prediction of turbofan engines are crucial to improving their operational safety and reliability. In order to improve data quality and mine key characteristic variables, this paper proposes a feature analysis method for turbofan engine time series data based on adaptive entropy value. First, the data is preprocessed to clean up outliers and missing data to ensure data integrity. Secondly, the correlation between sensor data is calculated using mutual information to mine potential correlation features, and the importance of each sensor parameter to engine life prediction is measured by conditional entropy. Subsequently, the adaptive entropy weight method is used to screen and optimize the feature variables to improve the quality of model training data. Finally, based on the GRU (Gated Recurrent Unit) deep learning model, experiments on the CMAPSS dataset show that this method can effectively improve the prediction accuracy and reduce the root mean square error (RMSE), verifying the effectiveness of the feature analysis method based on adaptive entropy value in turbofan engine life prediction.