Contribuição para monitoramento de roçamento em máquinas rotativas utilizando a técnica de emissão acústica

Abstract

Early detection of faults in rotating machinery is critical for maintaining operational reliability and efficiency across various industrial sectors. The Acoustic Emission technique has emerged as a promising method for structural health monitoring, particularly in identifying incipient defects such as light rubbing. Light rubbing faults, defined by contact between rotating and stationary surfaces, can accelerate mechanical degradation and result in catastrophic failures if not promptly detected. This thesis examines the use of the Acoustic Emission technique for detecting and characterizing rubbing faults in rotating machinery, with a focus on low-cost piezoelectric sensors, including lead zirconate titanate transducers and buzzers, integrated with machine learning algorithms for signal processing and classification. Experimental data were collected using a dedicated test rig under multiple operating conditions. Analyses employed both conventional signal processing methods, such as the Fast Fourier Transform and envelope analysis based on the Hilbert Transform, as well as machine learning-based techniques. The findings indicate that Acoustic Emission enables reliable identification of light rubbing faults even in noisy environments where traditional vibration analysis methods are limited. These results underscore the potential of Acoustic Emission as an effective and complementary approach for continuous condition monitoring of rotating machinery.