ニューラルネットワークの逆解法による主軸冷却油最適温度の推定 : 再学習の効果(機械要素,潤滑,工作,生産管理など)

概要

論文の詳細を見る
As higher precision machining is required, reducing thermal deformation of a machine tool becomes more important. In the previous report, a new method for estimating optimum temperature of cooling oil on a spindle was established. Inverse analysis using neural network was used for calculating optimum temperature, because neural network is suitable for controlling non-linear phenomenon. In this report, effect of relearning by inverse analysis using neural network was investigated. At first, in the experiment, thermal behavior of a bench lathe was measured for a few basic study data of the neural network. Next, the neural network was used to learn the relationship between the measured temperatures and the thermal deformation on the bench lathe. Then, optimum temperature for cooling oil on a spindle was calculated by inverse analysis using previous neural network, and oil temperature of the refrigerator was changed to the optimum temperature. At last, new neural network using the basic study data and the controlled data rebuild up, then the refrigerator for spindle of the bench lathe was controlled by inverse analysis using new neural network and was evaluated. It is concluded from the results that the proposed relearning was effective in order to reduce thermal deformation.
一般社団法人日本機械学会の論文
2003-03-25