V形切欠きを有する丸棒試験片の応力集中係数の推定法に対するニューラルネットワークの応用

抄録

We propose a neural network technique for estimating precisely the stress concentration factor of specimens with arbitrary configuration parameters, based on the finite number of values obtained discretely by Body Force Method. The validity of values estimated by the neural network technique is checked and confirmed by treating a problem of 60 degress V-notched cylindrical specimens under tension as an example. Finally, we make it clear how to select the rational training data necessary to train the neural network.