Statistical analysis of sloshing loads and performance of slosh-induced energy harvesting

Abstract

This study is the statistical analysis of sloshing loads and performance of slosh-induced energy harvesting. In associated with the subject, this thesis is largely composed of three topics. The first topic studied the sloshing load characteristics by the cylinder structure for the nuclear reactor of floating nuclear power plant. While there are cases in which internal structures are artificially installed to reduce sloshing, there are cases in which internal structures are installed for special purposes, such as nuclear reactors in the offshore nuclear power plant. A cylinder structure such as a nuclear reactor can be subjected to an impact load due to sloshing. For this case, evaluation of the peak pressure of the sloshing and examination of the phenomenon of internal flow are very important from the viewpoint of reliability design of the sloshing tank. In the first study, the impact of the cylinder on the shape of the sloshing flow, the magnitude of the peak pressure, and the velocity field at the moment of impact were analyzed. In addition, a probability model of the peak pressure on the tank wall was presented and compared with various extreme probability model when the cylinder is installed is presented. In addition to the model experiment, the probability model proposed in the model experiment was applied using the reactor model equipped with complex pipes in a large-scale sloshing tank. Through this, the Weibull distribution can be the most suitable extreme probability model for the reactor model under sloshing flow. The second topic is a study on predicting sloshing pressure using image-based deep learning. The wave profile immediately before impact was observed with a high-speed camera, and the sloshing pressure was measured at the same time to synchronize the image and pressure. By labeling the image and pressure that occurred right before the impact, the wave profile was learned through the residual neural network, and from this, this study was conducted to predict the regular, impulse, random, and peak pressure. As a result, the deep learning model established succeeded in predicting various sloshing pressures and showed excellent prediction performance. In particular, in the case of peak pressure, the normalization method has a great effect on the magnitude and profile of pressure. Moveover, the residual neural network well detected different types of wave images continuously coming in and predicted the corresponding pressure with the single algorithm. The third study is the energy harvesting of flexible sheet caused by sloshing. Sloshing has high wave energy right before impact, which is the energy dissipated after impact with the tank wall. In this paper, this study was conducted to convert wave energy caused by sloshing wave into electrical energy by installing a piezoelectric transducer on the flexible sheet, and the conditions for maximum energy harvesting were presented. Energy harvesting using the bending characteristics of the sloshing-based flexible sheet showed higher energy density than conventional devices using fluid flow, confirming the potential as a power supply device when sloshing occurs. From this, the wave induced by sloshing can be utilized as a sufficiently usable energy source as renewable energy.