Comparação de métodos de tratamento de sinais aplicáveis ao diagnóstico de defeitos em mancais de rolamento

Abstract

The choice of the right strategy for a plant maintenance is critical to the availability and reliability of its assets. The fact that most of the failures on rotary machines derive from damages on its rolling element bearings justifies the use of a method capable to detect and diagnose defects while incipient. For this reason, the present wok analyzed the diagnose capability of three different methods, among them the envelope analysis, plotted by many authors as the state of the art, and the local mean decomposition, LMD, and its optimized version, OELMD, when associated to an expert system specially designed for this. Other features, such the signal separation potential, the spectral quality and the computational consumption required, was also taken into account. The algorithms from several papers where translated to MatLab and tested with real vibration signals collected from rolling element bearings exposed to different kinds of damages. The comparison between the filter options to the envelope analysis pointed that time synchronous averaging, TSA, was better than the others in terms of cleansing quality and kurtosis amplification. In the case of the local mean decompositions, LMD and OELMD, this relevant content was revealed by the continuous decomposition of the signal into product functions or modes, which are the best representation of the original signal in terms of an envelope and a frequency modulated signal. In this sense, an intense mode mixing was observed in the answers provided by the LMD method, showing its inability to deal with complex and highly modulated signals. On the other hand, the OELMD method was capable to bring to the surface the impulsive features of the bearings previously masked by the more impulsive deterministic content of the signal. At the end the study showed that all three methods were equally assertive, although the hypothesis of the analyst dismissal is not possible while processing time and diagnostics quality are both required.