Wear debris monitoring of lubricant oil is an important method to determine the health and failure mode of key components such as bearings and gears in rotatory machines. The permittivity of lubricant oil can be changed when the wear debris enters the oil. Capacitive sensing methods showed potential in monitoring debris in lubricant due to the simple structure and good response. In order to improve the detection sensitivity and reliability, this study proposes a new coaxial capacitive sensor network featured with parallel curved electrodes and non-parallel plane electrodes. As a kind of through-flow sensor, the proposed capacitive sensor network can be in situ integrated into the oil pipeline. The theoretical models of sensing mechanisms were established to figure out the relationship between the two types of capacitive sensors in the sensor network. The intensity distributions of the electric field in the coaxial capacitive sensor network are simulated to verify the theoretical analysis, and the effects of different debris sizes and debris numbers on the capacitance values were also simulated. Finally, the theoretical model and simulation results were experimentally validated to verify the feasibility of the proposed sensor network.

No abstract

In addition to lubricating and cooling, aero-engine lubricating oil is also a transport medium for wear particles generated by mechanical wear. Online identification of the number and shape of wear particles is an important means to directly determine the wear state of rotating parts, but most of the existing research focuses on the identification and counting of wear particles. In this paper, a qualitative classification method of wear particle morphology based on support vector machine is proposed by using the wear particle capacitance signal obtained by the coaxial capacitive sensing network. Firstly, the coaxial capacitive sensing network simulation model is used to obtain the capacitance signals of different shapes of wear particles entering the detection space of different electrode plates. In addition, a variety of intelligent optimization algorithms are used to optimize the relevant parameters of the support vector machine (SVM) model in order to improve the classification accuracy. By using the processed data and optimized parameters, a SVM-based qualitative classification model for wear particles is established. Finally, the validity of the classification model is verified by real wear particles of different sizes. The simulation and experimental results show that the qualitative classification of different wear particle morphologies can be achieved by using the coaxial capacitive sensing network signal and the SVM model.