| 纪皓迪,马小兵.基于机器学习分类算法解析EIS数据的有机涂层性能评价方法[J].装备环境工程,2024,21(5):142-149. JI Haodi,MA Xiaobing.An Organic Coating Performance Assessment Method Based on Machine Learning Classification Algorithms and EIS Data[J].Equipment Environmental Engineering,2024,21(5):142-149. |
| 基于机器学习分类算法解析EIS数据的有机涂层性能评价方法 |
| An Organic Coating Performance Assessment Method Based on Machine Learning Classification Algorithms and EIS Data |
| 投稿时间：2024-03-22 修订日期：2024-04-23 |
| DOI：10.7643/issn.1672-9242.2024.05.016 |
| 中文关键词: 有机涂层 分类算法 机器学习 电化学阻抗谱 支持向量机 k最近邻 随机森林中图分类号:TG174 文献标志码:A 文章编号:1672-9242(2024)05-0142-08 |
| 英文关键词:organic coating classification algorithm machine learning electrochemical impedance spectroscopy Support Vector Machine k-Nearest Neighbor Random Forest |
| 基金项目: |
|
|
|
|
| 摘要点击次数: |
| 全文下载次数: |
| 中文摘要: |
| 目的 基于机器学习分类算法快速评估有机涂层的防腐性能。方法 通过实验室加速试验模拟涂层真实的退化过程，并根据测得的电化学数据，分析不同退化阶段的等效电路元件参数。随后，采用随机抽样方法获取大量数据，用于机器学习模型训练。通过对比支持向量机(SVM)、k最近邻(k-NN)和随机森林(RF)3种不同的机器学习算法，以及多种输入特征集训练的涂层性能分类器模型的准确率，分析最适合用于涂层性能快速评估的机器学习算法和电化学特征。结果 根据不同输入特征训练的k-NN和RF模型均表现出良好的预测效果，而SVM模型的预测效果相对较差。根据不同频率范围训练的分类器模型中，在低频区表现最佳，而在高频区表现较差。结论 基于阻抗虚部、虚部+实部和阻抗模值3种输入特征训练的RF分类器模型的预测效果最准确。不同频率区间内，低频区的阻抗特征更能准确表征涂层性能。 |
| 英文摘要: |
| The work aims to rapidly evaluate the corrosion resistance performance of organic coatings using machine learning classification algorithms. Laboratory accelerated tests were conducted to simulate the actual degradation process of coatings. The equivalent circuit parameters at different degradation stages were analyzed based on measured electrochemical data. Subsequently, a large amount of data were obtained for machine learning through random sampling. By comparing Support Vector Machine (SVM), k-Nearest Neighbors (k-NN), and Random Forest (RF) algorithms, as well as the accuracy of coating performance classifier models trained with various input feature sets, the most suitable machine learning algorithms and electrochemical features for rapid coating performance evaluation were analyzed. Classifier models trained with k-NN and RF models, demonstrated good predictive performance, while the SVM model showed relatively poorer predictive performance. Among the models trained with different frequency ranges, those trained with low-frequency data performed the best, whereas those trained with high-frequency data performed relatively worse. The RF classifier model trained with impedance imaginary part, imaginary part & real part, and impedance modulus as input features demonstrates the most accurate predictive performance. Within different frequency ranges, impedance features from the low-frequency range are more effective in accurately characterizing coating performance. |
| 查看全文 查看/发表评论 下载PDF阅读器 |
| 关闭 |
