Several details of the mechanism of gear lubrication are still in doubt in spite of many decades of study of this subject. The focus of our work is the investigation of the mechanism by which oil † temperature variations affect gear idle rattle, which requires an understanding of the distributions of lubricant and heat within a gearbox. This paper presents the findings of a study of lubricant flow in a simple model gearbox by means of CFD (Computational Fluid Dynamics) and its validation by a series of tests on a spur gear rig. The commercial CFD code Fluent is used to simulate the splash flow of lubricant, using the techniques of dynamic meshing and VOF (Volume of Fluid). Our model takes into account the effects on the distribution of gear lubricant of lubricant level and physical properties as well as rotational speed. The results demonstrate that the flow patterns are strongly influenced by all these variables. The predictions are validated by high-speed flow visualisations using high-resolution imaging in conjunction with a pulsed Cu-vapour laser light source and powerful white light source. The simulated fluid flows are in good qualitative agreement with the experimental visualisation. Estimates of the velocity of oil droplets from the images are compared with CFD predictions of the velocity in regions with high lubricant-phase-fraction resulting from lubricant splashing.

Post Operation Clean Out (POCO) is the process to remove hazardous materials and decommission nuclear facilities at the end of a nuclear plant’s lifetime. The introduction of Internet of Things (IoT) technologies in the environment, especially radio frequency identification (RFID), would improve efficiency and safety by intelligently monitoring POCO activities. In this paper, we present a passive material identification and crack sensing method developed for the integration of sensing and communication using commercial off-the-shelf (COTS) RFID tags, which is a long-term solution to material property monitoring under insulation for harsh environmental conditions. To validate the effectiveness of material identification and crack monitoring, machine learning techniques have been applied, and the feasibility of the study has been outlined. The result shows that the material identification can be achieved with traditional features and obtain improved accuracy with three-layer multi-layer neural networks (MLNN). In crack characterization, the tree algorithm based on traditional features achieves a reasonable accuracy, while three-layer MLNN is the best solution, which supports the efficiency of traditional feature extraction methods in specific applications.

No abstract