Prediction of Grain Size in the TiN Coating Using Artificial Neural Network

摘要

This paper presents a predicting model for the coating grain size in the machining process by considering the Artificial Neural Network (ANN) as the prominent technique for measuring thin film grain size. ANN technique ability for predicting grain size and machining process are highlighted. In order to investigate estimating the prediction value for grain size, a real machining experiment is utilized to validate and show the ANN technique modeling capability. Based on RSM center cubic design, 17 experimental data are collected. All data samples of tungsten carbide inserts are coated with TiN used physical vapour deposition process (PVD). A value of TiN coating grain size is measured using the atomic force microscopy (AFM). Feedforward backpropagation is selected as the ANN algorithm with trainlm, learngd, MSE, and logsig as the training, adaption learning, performance and transfer functions, respectively. Three parameters which are Argon gas pressure (Ar), Nitrogen gas pressure (N_2) and Turntable speed (TT) are represented as input nodes in the input layer. On the other hand, grain size node for the output layer. Eleven networks are designed and developed by using different structures and numbers of hidden layers and it's nodes. It was found that the 3-6-6-1 network structure of the TiN coated tool represented as optimal structure and gave the best ANN model in predicting the grain size value. This study concludes that the model for TiN coating grain size could be improved by modifying the structure of ANN as the number of hidden layers and its nodes, particularly for predicting the value of the grain size performance measure. As a result of the model prediction.