基于灰色模型和神经网络的铝合金腐蚀预测对比

采用NaCl溶液对铝合金试验件进行预腐蚀试验，产生腐蚀坑，获取了不同腐蚀时间下的腐蚀数据，然后进行疲劳加载试验。分别利用灰色模型和BP神经网络建立了腐蚀深度及疲劳寿命与腐蚀时间相关性的预测模型，对两种预测模型的精度进行了对比。研究发现，在缺乏足够统计数据的情况下灰色模型预测精度优于神经网络算法。

Comparative Study of Prediction Models of Aluminum Alloys Based on Gray Model and Artificial Neural Network

Aluminum alloy was tested through pre-corrosion in NaCl solution. The corrosion pits were detected to get corrosion damage data of different corrosion time. Corrosion fatigue test of the pre-corroded specimen was carried out. Gray prediction model and BP neural network algorithms were selected to establish predictive model of the relation between fatigue lives, corrosion depth, and corrosion time. The accuracy of both two prediction model were compared. The result showed that the prediction accuracy of gray prediction model is higher than BP neural network algorithm when the statistics is lack.