Fault diagnosis in electrical motors through vibration monitoring using Fiber Bragg Grating-based accelerometers
Nombre: LEANDRO CASSA MACEDO
Fecha de publicación: 17/10/2023
Junta de examinadores:
Nombre |
Papel |
|---|---|
| ANSELMO FRIZERA NETO | Coorientador |
| ARNALDO GOMES LEAL JUNIOR | Presidente |
| CAMILO ARTURO RODRIGUEZ DIAZ | Examinador Interno |
| RUI MIN | Examinador Externo |
Sumario: Structural Health Monitoring (SHM) techniques have been explored in fault and damage diagnosis in structures and machines. These techniques are explored within the scope of Industry 4.0 and Smart Cities, where the data provided by SHM techniques
are used in the development of predictive and preventive maintenance plans, avoiding catastrophic failures, reducing machinery downtime, and providing more security in the cities. Sensor development plays an important role in this scenario since these devices are responsible for turning physical measurements into data that are capable of being processed to provide data-based decisions in industrial processes and city management. Different types of sensors are developed to attend to industrial requirements, such as thermocouples for temperature measurements, accelerometers for acceleration measurements, and strain gauges for strain measurements. In this context, optical fiber sensors can offer some advantages for sensor applications: they can be immunity to electromagnetic interference (ideal for industrial harsh environments), can be easily embedded into structures since they are thin and flexible, can be multiplexed (i.e. produce multiple sensors in the same optical fiber cable), and can combine sensing and data transmission over long distances applications using the same optical fiber cable. In this work, a Fiber Bragg Grating-based accelerometer design is reported for machinery fault diagnosis. Different geometries are analyzed as candidates for developing FBG-based accelerometer projects. Through analytical models, the flexible hinge structure was selected based on the sensitivity and natural frequency features to attend to the project requirements. The geometric dimensions are then selected by a multi-objective optimization procedure, in which a variety of combinations of geometric parameters are evaluated with respect to sensitivity and natural frequency. This procedure served as an efficient tool for varying different geometric parameters to find combinations that maximize sensitivity and natural frequency. Four structures are selected to compose this work and, before fabrication, they are analyzed using a Finite Element Modal Analysis. These results were compared to the analytical model results, implying relative errors of 23%, 33%, 14%, and 6% for accelerometers 1, 2, 3, and 4, respectively. These
errors are related to idealizations assumed and neglected effects in the analytical models. The sensors were then fabricated and characterized. The experimental natural frequencies were 607.8 Hz, 366.7 Hz, 294.7 Hz, and 236.5 Hz for accelerometers 1, 2, 3, and 4, respectively. The experimental sensitivities are characterized by the exciting frequencies of 17 Hz, 35 Hz, and 50 Hz. For 17 Hz, the experimental sensitivities were 180 pm/g, 690 pm/g, 380 pm/g, and 400 pm/g, for accelerometers 1, 2, 3, and 4, correspondingly. For 35 Hz, the experimental sensitivities were 150 pm/g, 510 pm/g, 290 pm/g, and 230 pm/g, for accelerometers 1, 2, 3, and 4, respectively. For 50 Hz, the experimental sensitivities were 120 pm/g, 410 pm/g, 150 pm/g, and
160 pm/g, for accelerometers 1, 2, 3, and 4, respectively. These sensors were applied in fault diagnosis experiments for 9 fault conditions, where the results were compared and validated by a commercial piezoelectric accelerometer. The comparison
between the identified peaks by the FBG-based accelerometers with the results obtained by the PZT-based accelerometers can be used to estimate an average relative error. For the FBG-based accelerometer 2, the relative errors are 0.48%, 0.62%, 0.50%,
0.32%, 0.76%, 0.26%, 0.39%, 0.40%, and 0.48%, for fault conditions 1, 2, 3, 4, 5, 6, 7, 8, and 9, respectively. For the FBG-based accelerometer 3, the relative errors are 0.27%, 0.72%, 0.39%, 0.22%, 0.95%, 0.29%, 0.29%, 0.40%, and 0.85%, for fault conditions 1, 2, 3, 4, 5, 6, 7, 8, and 9, respectively. For all cases, the FBG-based accelerometer iii frequency vibration spectra were similar to the piezoelectric accelerometer measurements, and it was concluded that the projected accelerometers in this work identified
correctly the vibration pattern in all fault conditions.
