ARTIFICIAL NEURAL NETWORK FOR AUTOMATIC PREDICTION OF THE SURFACE FINISHING VIA CLASSIFICATION OF THE SURFACE TEXTURE

In this study, the effects of the machining type on the notch of a specimen, made of 7075-T6 aluminum alloy, have been investigated. The CNC machining and waterjet cutting are used to machine specimens for in fatigue testing. The surface texture characterization, resulting from machining, is examined by using an optical metrology device (Alicona). The tested surface texture is characterized by the average height, Ra, over a small selected range of data. The effect of the stress concentration was studied by designing the surface texture using a computer-aided design (CAD) tool. It is observed that the waterjet cutting produces very rough surfaces, hence increasing the fatigue stress concentration. In addition, this study presents an automated prediction for the machining type using the tools of artificial intelligence (AI), where a neural network (NN) is applied to predict the machining type, which is based on the extracted surface textures, derived from the measurement data. The classification methodology, which uses a NN model, has been built based on the concept of pattern recognition. Results of this study show that the NN model is capable of successfully predicting the type of surface roughness with (up to) 90% accuracy.