トライボロジスト
Online ISSN : 2189-9967
Print ISSN : 0915-1168
ISSN-L : 0915-1168
最新号
特集・カーボンニュートラルに貢献する転がり軸受の技術動向
選択された号の論文の15件中1~15を表示しています
会告
目次
連載・トライボロジーを語る
特集・カーボンニュートラルに貢献する転がり軸受の技術動向
解説
トライボロジー・ナウ・トライボエピソード ―奨励賞受賞―
学術論文
  • 畠中 清史, 内田 渓太郎
    2024 年 69 巻 10 号 p. 690-707
    発行日: 2024/10/15
    公開日: 2024/10/15
    [早期公開] 公開日: 2024/08/10
    ジャーナル フリー

    Thermohydrodynamic lubrication (THL) models have been applied to predict the performances of journal bearings that support industrial rotating machineries. This article aims at deriving a model expression that can easily predict the maximum bearing temperature in the THL database of cylindrical journal bearing with two-axial oil grooves, by applying deep learning (DL). The expression is given only four dimensionless bearing design variables. The DL model consists of a multi-layer perceptron with a preprocessing layer, four hidden layers, an output layer and a postprocessing layer. All nodes in adjacent layers are fully connected. Logarithmic function is employed as preprocessing function for the bearing design variables and the maximum bearing temperature, and inverse tangent function as activation function. Six nodes are arranged in each hidden layer. The dataset with the maximum bearing temperature above 0.2 is selected for the training data. Learning is continued until the maximum relative errors for the training and the validation data simultaneously fall within the criterion value. The model expression is shown to be capable of easily predicting the maximum bearing temperature quickly and with high generalization performance and is concluded to be quite useful in significantly reducing the time required to THL design of the bearing.

会のページ
feedback
Top