Methodical Evaluation of Sensor Positions for Condition Monitoring of Gears
DS 91: Proceedings of NordDesign 2018, Linköping, Sweden, 14th - 17th August 2018
Year: 2018
Editor: Ekströmer, Philip; Schütte, Simon and Ölvander, Johan
Author: Martin, Georg; Vogel, Sven; Schirra, Tobias; Vorwerk-Handing, Gunnar; Kirchner, Eckhard
Series: NordDESIGN
Institution: Technische Universität Darmstadt
ISBN: 978-91-7685-185-2
Abstract
With the increasing digitalization of mechanical engineering and the development of mechanical systems into cyber-physical systems, predictive maintenance and condition monitoring have become more widespread in technical applications. However, the fault diagnosis and especially the prognosis of remaining useful lifetime is still not accurate in many applications. This paper presents a methodical approach for evaluating and choosing measurands and sensor positions in order to improve condition monitoring systems by enhancing the quality of the measurement data. The underlying concept of sensor-based condition monitoring systems is that faults affect measurable properties of the machine, enabling the deduction of the fault status by analyzing the measured data. Most current systems use quantities that are easy to measure, like vibration at the machine housing, and successively apply advanced data processing methods for diagnosis and prognosis. For many failure mechanisms, these commonly used measurands are too far away from the actual fault location, e.g. a tooth root crack in a spur gear transmission. This leads to a loss of information between the fault location and the sensor: The property of interest (depth of the tooth root crack) is transformed into a variation of the contact force and mechanical vibration. Thus, the vibration can be seen as a mechanical signal, which is masked by disturbances, attenuated and modified along the transmission path from its source to the sensor. Therefore, the information about the fault, which is conveyed by the vibration signal, is more accurate when the sensor is positioned close to the fault location. The consequential problem for the developer of condition monitoring systems is to choose measurands and sensor positions in order to maximize the quality of the gathered data. A methodical approach for evaluating measurands and sensor positions is presented and explained in this paper using the example of gear condition monitoring. First, possible measurands and sensor positions are explained. Then, sensor positions for vibration-based condition monitoring are evaluated by estimating and measuring the vibration transfer function between the fault location (gear tooth) to various sensor positions, both on the shaft and on the housing. The results show that the bearings, due to their compliance, isolate vibration and thus decrease the information content of vibration measurements at the housing, compared to measurement on the shaft. Therefore, it is proposed to develop new sensor concepts that allow cyber-physical systems to reach deep into machines and gather relevant information closer to the actual point of interest, in this case the fault location.
Keywords: Big Data/ Digital Design, Cyber-Physical Systems, Condition Monitoring, Predictive Maintenance, Sensor Integration, Tooth Root Crack, Bearings