Prediction of acid salt spray corrosion experiment of 2024 aluminum alloy based on RBF neural network

Abstract

With the increasingly serious industrial pollution, aircraft structural materials are inevitably exposed to acidic environment during the service. 2024 aluminum alloy was used for acid salt spray experiment under different conditions. The pH values of salt spray experiment were set as 2, 3, 5, the salt spray concentration was 25 g/L, 50 g/L, 75 g/L, and the corrosion time was 24 h, 48 h and 72 h respectively. Combine radial basis function (RBF) neural network with orthogonal experiment design, different experimental conditions were selected as the learning sample set of the neural network, and the results of orthogonal test are analyzed by range analysis. The results show that RBF combined with orthogonal experimental design can accurately predict the corrosion rates under arbitrary experimental conditions, reduce the number of experiment, and improve the prediction accuracy. The prediction result of using both the orthogonal group and supplementary peak group as the sample set is better than that of using only orthogonal group as the sample set. The results of range analysis show that the most important factor affecting the mass loss per unit area of 2024 aluminum alloy is the pH value of the solution, followed by the concentration of salt spray and the corrosion time is the least.

Keywords

• aluminum
• acid spray experiment
• orthogonal experiment
• RBF neural network