VIBRATION SIGNAL ANALYSIS AND MEASUREMENT METHOD AND DEVICE (V-SAM)

Abstract

The invention relates to the V-SAM device (Vibration Signal Analysis and Measurement), an advanced, portable vibration meter designed for monitoring and analyzing vibration signals from industrial machinery, and the system for this vibration meter. In other words, the invention incorporates specialized modules for accurately diagnosing faults in rotating mechanical components and analyzing the condition of bearings, gearboxes, and induction electric motors, using intelligent algorithms and advanced analysis techniques. A signal processing method and monitoring system for detecting periodic fault traces in vibration signals of rotating machinery is described. The invention presents a periodicity extraction framework based on a modified form of the Mean Magnitude Difference Function (AMDF) structured to improve the detectability of weak, irregular, or partially masked periodic structures in noisy measurement environments. In one aspect, a vibration signal obtained from rotating machinery is conditioned and processed to calculate a representation of the mean magnitude difference that characterizes the temporal periodicity associated with impulsive mechanical events. The resulting delay domain representation is processed to eliminate slowly changing trends and suppress short-delay components associated with broadband noise. The processed sequence is then transformed so that periodic minima are mapped to distinct peaks, thus providing a better definition of periodic structures related to mechanical failures. The improved periodicity representation is analyzed in more detail in the spectral domain to obtain a frequency-domain representation of the underlying periodic behavior. Statistical estimation of background spectral noise is performed using robust statistical estimators resistant to outliers. Based on predicted noise characteristics, an adaptive spectral mask is generated to selectively emphasize spectral components associated with periodic fault signatures while suppressing broadband noise and irrelevant spectral components. The method also derives a periodicity mask that identifies the synchronous and asynchronous periodic spectral components associated with mechanical excitation, independently of rotational speed synchronization. Then, characteristic failure frequencies can be extracted from the improved spectral representation, and a machine condition indicator or health index can be calculated as a function of the relative energy associated with the identified periodic components. The described framework can be implemented as a signal processing method, an embedded monitoring system, or a computer-implemented program for condition monitoring, predictive maintenance, and fault diagnosis of industrial rotating equipment.