石材研削におけるダイヤモンド・ソーブレードの確率論的性能予測 : ニューラルネットワークによる破砕確率,推移確率,脱落確率の推定

抄録

In order to predict the performance of the diamond saw blades during grinding, an analytical method with applying the theory of Marcov process was suggested in previous paper. This theory, which basically have the probability components consisting of initial shape distribution, fracture, transition, dislodgement and appearance, enable us to predict variation of edge shape distribution on wheel surface with grinding time. However, these probabilities had to be obtained directly from the experiment. Therefore, this method inevitably follows new experiment and measurement whenever experimental conditions change. It remains a question how to decide the probabilities efficiently for more practical method. In this study, estimation of these probabilities is attempted by the neural network. As a result, it is found that reasonable values of the probabilities can be obtained from input data of grinding condition given. Hence, applying the neural network can develop the theory proposed for more practical method.