The Application of Surrogate Models in the Study of the Fluid-Film Lubrication

• Main Authors: Kuo-Chih Huang, 黃國誌
• Other Authors: Nenzi Wang
• Format: Others
• Language: zh-TW
• Published: 2006
• Online Access: http://ndltd.ncl.edu.tw/handle/60815801259115423824

碩士 === 長庚大學 === 機械工程研究所 === 94 === In this study, an artificial neural network and a Kriging metamodel are applied to replace the mathematical model of the fluid-film lubrication. The main purpose is to evaluate the characteristics of the two metamodels for replacing computationally expensive lubrication analyses in tribology. The slider bearing performance obtained in this study is computed from a thermohydrodynamic lubrication model which is represented by the coupled Reynolds equation and a 3-D energy equation. The two design variables are the slope the slider bearing and the aspect ratio of a given area. These two equations are solved by successive over-relaxation and alternating direction implicit methods, respectively. The sampling points are selected symmetrically in the design space, and the validation points are selected both symmetrically and randomly. The criteria for the assessment of the metamodels are maximum absolute error, minimum error, maximum error percentage and root mean square error. In this study, we found that the two metamodels could be used to replace the mathematical model and help reducing the computation time.