ARTIFICIAL NEURAL NETWORK MODEL FOR SURFACE ROUGHNESS PREDICTION IN CNC MILLING MACHINING

• Main Author: Amir, Afrizal
• Format: Theses
• Language: Indonesia
• Online Access: https://digilib.itb.ac.id/gdl/view/53602
• Institution: Institut Teknologi Bandung

Surface finish is an important factor in evaluating the quality of products. Many factors influence the surface roughness of the product; one of them is cutting parameter. The aim of this research is to create an artificial neural network (ANN) model for surface roughness prediction in milling process. Cutting parameters of spindle speed, depth of cut, feed rate, and cooling fluid were studied in this research. This studies point out the usage of cooling liquid as an input parameter. Feedforward-backpropagation algorithm will be used in the development of ANN structure. Levenberg-Marquardt backpropagation training function and Gradient descent with momentum and adaptive learning rate backpropagation training function were selected in the study. Furthermore, hyperbolic tangent sigmoid transfer function and sigmoid transfer function were employed as activation function. Testing and verification used data outside of experiments were carried out. The performance of each model was evaluated by measuring mean squared error (MSE) of each model. The experiment result showed that optimum cutting parameter to produce best surface quality (Ra = 0.06 ?m) is combination of using cooling fluid-high spindle speed-low depth of cut-low feedrate. The ANN analysis results the 4-2-1 structure with the Levenberg-Marquardt backpropagation training function and the sigmoid activation function was the most accurate model with 0.004, 0.0017, and 0.0077 MSE values for training, testing, and verification respectively. Furthermore, realtime surface roughness prediction during the cutting process using ANN method can be investigated in the future