| 引用本文: | 李昭毅,孙虎元,蔡振宇,孙立娟.基于Sine-SSA-BP人工神经网络的腐蚀速率预测研究[J].海洋科学,2024,48(8):17-28. |
| |
|
| |
|
|
| 本文已被：浏览 8次 下载 1次 |
码上扫一扫！ |
|
|
| 基于Sine-SSA-BP人工神经网络的腐蚀速率预测研究 |
|
李昭毅1,2,3,4, 孙虎元1,2,4, 蔡振宇1,2,3,4, 孙立娟1,2,4
|
|
1.中国科学院海洋研究所海洋关键材料重点实验室, 山东 青岛 266071;2.中国科学院海洋研究所海洋环境腐蚀与生物污损重点实验室, 山东 青岛 266071;3.中国科学院大学, 北京 100049;4.中国科学院海洋大科学研究中心, 山东 青岛 266071
|
|
| 摘要: |
| 海洋工程用钢广泛应用于海洋资源开发; 然而, 在海洋环境中, 由于海洋环境复杂, 钢的腐蚀速度大幅加快。为了评估其使用寿命, 需要准确地预测钢的腐蚀速率。挂片实验法费时费力, 经验模型预测虽然可以快速预测, 但因海洋中影响腐蚀的因素较多, 准确度较差。本文介绍了一种机器学习方法, 即反向传播(BP)神经网络金属腐蚀速率预测模型。本研究创新性地将Sine混沌映射与麻雀搜索优化算法(SSA)引入腐蚀速率预测模型中, 并利用2022年采集到的海洋环境要素和腐蚀速率数据导入模型进行训练预测。结果表明, SSA-BP和Sine-SSA-BP神经网络金属腐蚀速率预测模型的误差远低于BP神经网络腐蚀速率预测模型。经过充分的训练和学习, 当预测样本数量由5至30逐渐增加时, Sine-SSA-BP预测模型的平均MAPE值为3.500 2%, SSA-BP模型的平均MAPE值为6.090 0%。 |
| 关键词: 海洋腐蚀 BP人工神经网络 麻雀搜索优化算法 预测精度 |
| DOI：10.11759/hykx20240312001 |
| 分类号:TP301.6 |
| 基金项目:国家自然科学基金项目(41476067) |
|
| Corrosion rate prediction based on Sine-SSA-BP artificial neural network |
|
LI Zhaoyi1,2,3,4, SUN Huyuan1,2,4, CAI Zhenyu1,2,3,4, SUN Lijuan1,2,4
|
|
1.Key Laboratory of Advanced Marine Materials, Institute of Oceanology, Chinese Academy of Sciences, Qingdao 266071, China;2.Key Laboratory of Marine Environmental Corrosion and Bio-Fouling, Institute of Oceanology, Chinese Academy of Sciences, Qingdao 266071, China;3.University of Chinese Academy of Sciences, Beijing 100049, China;4.Center for Ocean Mega-Science, Chinese Academy of Sciences, Qingdao 266071, China
|
| Abstract: |
| Steel is widely used in marine engineering for marine resource development. However, due to the complex marine environment, the corrosion rate of steel is significantly accelerated. Thus, accurately predicting the rate of steel corrosion is crucial to determining its service life. The hanging plate experimental method is cumbersome and tedious to perform. Moreover, as several elements influence corrosion, empirical models can make quick but often imprecise predictions. Herein, a back propagation (BP) neural network-based model for predicting metal corrosion rate—a machine learning technique—is presented. This study creatively incorporated the sparrow search optimization algorithm (SSA) and sine chaotic mapping into the model. The model was then trained using corrosion rate data from 2022 and the gathered elements of the maritime environment. The findings revealed that compared to the BP neural network models, the SSA-BP and Sine-SSA-BP neural network models have substantially smaller errors while predicting metal corrosion rates. Following adequate training and learning, as the number of predicted samples progressively rose from 5 to 30, the average mean absolute percentage error value of the Sine-SSA-BP prediction model was 3.500 2%, and that of the SSA-BP model was 6.090 0%. |
| Key words: marine corrosion BP artificial neural network sparrow search optimization algorithm prediction accuracy |
|
|
|
|
|
|