将神经网络用于金属土壤腐蚀研究，利用神经网络的学习特征和高度的非线性特征，以土壤理化性能、腐蚀时间、Ａ３钢在土壤腐蚀试验１、２、８个月的腐蚀数据作为网络训练样本，对土壤中埋片２４个月的Ａ３钢腐蚀速率进行预测，并对结果进行了分析。
Artificial neural network, possessing learning and non-linear character, was applied to study corrosion of mild steel in soil. A neural network typically consists of many simple neurons like processing elements called "cell" or "nodes" that interact with other cells using numerical weighted connection. In this study,a neural network with 6-10-1 structure, namely 6 input nodes, 10 hidden layer nodes, 1 output node, was used. The learning algorithm was BP (Back-Propagation) algorithm. Corrosion tests of mild steel in soil were carried out with orthogonal test method, in which five corrosion factors, namely pH value,Cl-, H2O,SO_4~2- and Fe2+ content, were considered and test data of sixteen groups were obtained. These data were used as sample set to train neural network. The inputs of neural network were the five corrosion factors and test duration, the output of neural network was corrosion rate of steel in soil.The research results showed that soil corrosion rates of mild steel in 24 months could be predicted by the trained artificial neural network, and were basically in agreement with experimental data.
参考文献
| [1] | |
| [2] | |
| [3] |
- 下载量()
- 访问量()
- 您的评分：
-
10%
-
20%
-
30%
-
40%
-
50%