Optimizing 5-axis sculptured surface finish machining through design of experiments and neural networks

Five axis machining and CAM software play key role to new manufacturing trends. Towards this direction, a series of 5 axis machining experiments were conducted in CAM environment to simulate operations and collect results for quality objectives. The experiments were designed using an L27 orthogonal array addressing four machining parameters namely tool type, stepover, lead angle and tilt angle (tool inclination angles). Resulting outputs from the experiments were used for the training and testing of a feed-forward, back-propagation neural network (FFBP-NN) towards the effort of optimizing surface deviation and machining time as quality objectives. The selected ANN inputs were the aforementioned machining parameters. The outputs were the surface deviation (SD) and machining time (tm). Experimental results were utilized to train, validate and test the ANN. Major goal is to provide results robust enough to predict optimal values for quality objectives, thus; support decision making and accurate machining modelling.