Automatic selection of the WPT decomposition level for condition monitoring of rotor elements based on the sensitivity analysis of the wavelet energy

Vibration signals are a widely used technique for machine monitoring and fault diagnostics. However, it is necessary to select a suitable pattern that represents the condition of the machine. Wavelet Packet Transform (WPT) provides a high potential for pattern extraction. Several factors must be selected and taken into account in the wavelet transform application such as the level of decomposition, the suitable mother wavelet, and which frequency bands (obtained from the decomposition process) contain the necessary information for the diagnosis system. The selection of the parameters commented above is a complex task that depends on many factors. Most of the works found in the literature select these factors based on experience, graphical methods, or trial and error methods. In this work, a method based on the relative wavelet energy is developed in order to automatically select the parameter defined by the WPT. The selection allows for the efficient extraction of the most suitable patterns for a later classification and fault detection process. In order to prove the soundness of the method proposed, a Jeffcott rotor model with four crack levels will be developed, and the vibratory signals provided by this model will be used for the monitoring condition.

Automatic selection of the WPT decomposition level for condition monitoring of rotor elements based on the sensitivity analysis of the wavelet energy Articles