Dynamically adjusted normalized multi-scale symbolic dynamic entropy for fault diagnosis of rotating machinery in strong noise

Noise is inevitable in practical industrial applications. The weak early fault characteristics of rotating machinery lead to serious performance degradation of existing feature extraction methods under noise interference. In view of this, this paper proposes a feature extraction method based on dynamically adjusted and normalized multi-scale symbolic dynamic entropy (DNMSDE). Firstly, the issue of entropy deviation caused by traditional methods is addressed using a coarse-grained method of mean standard deviation normalization. Subsequently, a dynamic adjustment method is introduced to improve the state mode probability based on the current position and the maximum embedding dimension in the time series, thereby better capturing the dependency of the sequence. Outlier detection in the time series is performed using the Z-score method to eliminate the influence of anomalous data. The proposed method is experimentally validated using simulated signals from three different datasets. The experimental results indicate that the proposed method can effectively extract signal features in a strong noise background and achieve higher diagnostic accuracy.