Springback optimization for CNC tube bending machine based on an artificial neural networks (ANNs)

Predicting the springback angle has become major production problem among tube benders. Springback is where on a mandrel-less rotary draw bending tends to bounce back after being bent when clamps are released. Accurately predicting crucial for effective bending. Machine learning (ML), popular prediction approach, was applied functions such as or function approximation, pattern classification, clustering, and forecasting. To achieve this, values from 27 experiments were collected used input into artificial neural networks (ANNs) in one area of ML. This research conducted study optimization ASTM A-210 Gr. A1 seamless with an outside diameter 44.45 mm, using 4 factors Wall Thickness, Bending Radius, Dwell Time, Angle. The results showed that all significantly influence process; different methods analyzed by comparing activation functions. optimal network architecture 4-98-1; these achieved Sigmoid function, giving lowest mean squared error (MSE) = 0.001892. resulting coefficient determination (R2 ) 99.42%, ReLU R2 98.99%, TanH 98.53%, Identity which 79.53%. It also found best regression equation, 82.32%, better than 65 neurons 79.53%, 2.79% difference favor equation.