| 李淮,张晓东,张传鸿,陈心欣,赵雪茹,揭敢新.基于数据驱动的汽车部件温度预测模型研究[J].装备环境工程,2023,20(2):102-109. LI Huai,ZHANG Xiao-dong,ZHANG Chuan-hong,CHEN Xin-xin,ZHAO Xue-ru,JIE Gan-xin.Temperature Prediction Model of Auto Parts Based on Data Driven[J].Equipment Environmental Engineering,2023,20(2):102-109. |
| 基于数据驱动的汽车部件温度预测模型研究 |
| Temperature Prediction Model of Auto Parts Based on Data Driven |
| |
| DOI：10.7643/issn.1672-9242.2023.02.014 |
| 中文关键词: 大数据分析 神经网络 机器学习 汽车自然暴露试验中图分类号:U467.1+3 文献标识码:A 文章编号:1672-9242(2023)02-0102-08 |
| 英文关键词:big data analysis neural network machine learning outdoor exposure test of vehicle |
| 基金项目: |
|
|
| Author | Institution |
| LI Huai | China National Electric Apparatus Research Institute Co., Ltd, Guangzhou 510663, China |
| ZHANG Xiao-dong | China National Electric Apparatus Research Institute Co., Ltd, Guangzhou 510663, China |
| ZHANG Chuan-hong | Nanjing University of Aeronautics and Astronautics, Nanjing 210016, China |
| CHEN Xin-xin | China National Electric Apparatus Research Institute Co., Ltd, Guangzhou 510663, China |
| ZHAO Xue-ru | China National Electric Apparatus Research Institute Co., Ltd, Guangzhou 510663, China |
| JIE Gan-xin | China National Electric Apparatus Research Institute Co., Ltd, Guangzhou 510663, China |
|
| 摘要点击次数: |
| 全文下载次数: |
| 中文摘要: |
| 目的 基于某汽车在中国吐鲁番地区自然暴露的部件温度变化试验数据，预测该车在美国凤凰城地区气象环境下的汽车部件温度变化。方法 把汽车部件的温度作为输出变量，提取影响汽车部件温度变化的关键特征(试验时间、大气温度、太阳辐照)作为输入变量，同时运用公式对不同纬度地区部件受到的太阳辐照进行修正，以消除地理位置的影响。利用Python等软件构建机器学习模型，用吐鲁番试验数据训练模型，然后预测该车部件在美国凤凰城地区的温度变化。结果 梯度提升机模型具有良好的泛化能力和预测精度，其预测值与实际值的平均绝对误差均在3.3°以内，拟合优度R2均大于0.90。BP神经网络和随机森林算法模型也具有较好的预测精度。结论 利用汽车在我国试验站点进行的自然暴露试验数据，可以预测该汽车部件在国外相似地区气象条件下的温度变化。该研究对于依据汽车部件在我国的自然暴露试验结果预测其他国家相似地区自然环境下汽车部件的温度变化具有一定的指导意义。 |
| 英文摘要: |
| The work aims to predict the temperature changes of the parts of a car in the Phoenix area of the United States based on the test data of the temperature changes of the parts of the car naturally exposed in Turpan, China. With the temperature of the auto parts was taken as the output variable, the key features that affect the temperature changes of the auto parts (test time, atmospheric temperature, solar radiation) were extracted as the input variables. At the same time, a formula was used to correct the solar radiation of parts in different latitudes to eliminate the effects of geographic location. Software such as python were used to build a machine learning model. The test data in Turpan were used to train the model, and then the temperature change of the auto parts in the Phoenix area of the United States was predicted. The prediction results showed that the gradient boosting algorithm model had good generalization ability and prediction accuracy. The average absolute error between the predicted value and the actual value was within 3.3 degrees, and the goodness of fit R2 was greater than 0.90. The BP neural network and random forest algorithm models also had good prediction accuracy. Using the natural exposure test data of a car at a test site in my country could predict the temperature changes of auto parts under the meteorological conditions in similar regions abroad. The research in this work has certain guiding significance for predicting the temperature changes of auto parts under the natural environment of similar regions in other countries based on the results of the natural exposure test of auto parts in China. |
| 查看全文 查看/发表评论 下载PDF阅读器 |
| 关闭 |
|
|
|
