APSO-BPNN模型在滨海环境中铁质材料腐蚀速率预测中的应用

Application of APSO-BPNN Model in Prediction of Iron Material Corrosion Rate in Coastal Environment

摘要: 针对滨海复杂环境中铁质材料腐蚀速率预测的问题,利用自适应粒子群优化（APSO）算法对反向传播神经网络（BPNN）中的权重和阈值进行优化,构建了一种APSO-BPNN模型,以提高铁质材料在滨海环境中腐蚀速率预测的准确性。基于暴露试验数据,对比了APSO-BPNN模型与传统BPNN模型的预测效果。结果表明：APSO-BPNN模型在训练集上的决定系数R²提高了23.65%,其在测试集上的R²达到0.925 8,平均绝对误差（MAE）、平均绝对百分比误差（MAPE）和均方根误差（RMSE）分别下降至11.55、22.26%和14.43。

Abstract: Aiming at the problem of corrosion rate prediction of iron materials in coastal complex environment, an adaptive particle swarm optimization (APSO) algorithm was used to optimize the weights and thresholds in back propagation neural network (BPNN), and an APSO-BPNN model was constructed to improve the accuracy of corrosion rate prediction of iron materials in coastal environment. Based on the exposure experimental data, the predictive effects of APSO-BPNN model were compared with those of traditional BPNN model. The results showed that the APSO-BPNN model enhanced the determination coefficient R² by 23.65% on the training set. Its R² on the test set reached 0.925 8, and the mean absolute error (MAE), mean absolute percentage error (MAPE) and root mean square error (RMSE) decreased to 11.55, 22.26 % and 14.43, respectively.