人工知能アルゴリズムを用いた噛合型同方向回転二軸押出機スクリュデザインの自動最適化

抄録

Deep learning using an artificial intelligence algorithm was carried out in order to automatically determine the screw configuration of an intermeshing co-rotating twin screw extruder, and a system development was conducted to obtain an optimum output value for the specified extrusion conditions from the learning results. In deep learning, supervised learning was adopted that gives extrusion conditions and machinery configuration to the input layer and sets the analysis results and experiment results to the output layer. After performing the deep learning, judgment processing was carried out under the conditions of complete melting extrusion at just 200℃ as designation conditions, and four kinds of screw design were outputted. As a result of experiment verification of these four screws, it was confirmed that they satisfy the specified value with high accuracy. This suggested that it is possible to determine the screw configuration with deep experience and expertise by utilizing the artificial intelligence algorithm.