作者:
鲁庆， 穆志纯
关键词:
摘要:
To study the corrosion pattern of material in natural environment, a corrosion rate model was established with carbon steel corrosion data in soil as the study object. A new algorithm was proposed in modeling which is based on boosted regression trees. For parameter selection in the condition of small sample data, the new algorithm improved the original algorithm by using 蔚 insensitive loss function and dynamic shrinkage coefficient. The new algorithm was compared with typical algorithms such as neural network and support vector regression (SVR). The simulation and test results show that the improved BRT algorithm has great robustness for solution of data problems including high dimension, missing value and high noise, and it is suitable for processing small sample data. The model established with this algorithm can precisely describe and predict the corrosion rate of carbon steel in soil, and also be used for exploratory analysis of corrosion affecting factors and the interaction between these factors.
在线下载