Manutenção preditiva com sistema de identificação de vibrações em vídeo usando ampliação de movimento

Abstract

Predictive maintenance is the most suitable for industries, as it monitors the health of machines and provides the least amount of downtime in the production line. One of the monitored signals is the vibration level of the machines. The instruments most used for this measurement are vibration sensors, meters and analyzers, which mostly need physical contact with the monitored equipment. However, these instruments are difficult to be attached to old machines, depend on wires and can malfunction, generating wrong measurements. Solutions that measure the level of vibration without contact are expensive. Therefore, the objective of this research is to implement and develop a computational solution to measure, in real time and without contact, the vibrations of electric motors in multiple areas of interest. This objective is achieved by proposing the Video Vibration Identification System using Motion Magnification (SIVVAM), in which the main contributions are the identification of non-contact motor vibration, real-time processing, definition of two areas of interest to measure vibration simultaneously, which provides greater mobility and processing speed to the proposed system in relation to existing solutions. The proposed solution can assist in predictive maintenance by alerting when the vibration level is above the standard indicated by the ISO 10816-3 standard.