Design of Neural Network Model and its Application to Coating Performance Prediction

Abstract

Abstract 316L stainless steel has excellent mechanical properties and good corrosion resistance. It is used as a medical implant material, but it is susceptible to pitting corrosion to precipitate harmful ions and lead to the failure of the material. To solve the above problems, the surface modification method is used to solve it, and it is now widely accepted. Among them, magnetron sputtering technology has been widely used because of its advantages such as good adhesion and other advantages. The parameters of the RF magnetron sputtering process have a great influence on the performance of the coating, and the influence of each parameter has a non-linear mapping ability. The artificial neural network is used to build a model to predict the performance of the coating. It is highly targeted and practical. The establishment of artificial neural network is a prediction model from sputtering process parameters to coating performance, which can save pre-research time and improve work efficiency. The establishment of artificial neural network uses the strong nonlinear mapping ability of artificial neural network, combined with the measured value of Hf-based coating performance to establish a model to predict the effect of Hf-based coating sputtering process parameters on the performance of the coating. The simulation results indicate the prediction The error between the value and the measured value meets the BP network design requirements ≤0.3.