摘要
通过研究聚碳酸酯(PC)在西藏自治区(西藏)10个典型大气站点自然暴露试验1年的老化行为,以色差作为指标评估西藏地区PC服役大气环境严酷度时空分布规律.采集了西藏10个典型大气站点8类环境因素数据年均值(2021年4月—2022年3月),分析了西藏地区气候环境特点及气候分布区域,为环境严酷度评价提供准确输入.通过在10个站点开展自然环境试验,研究发现PC老化过程中表观失光率、色差逐渐上升,力学性能如拉伸强度、拉伸断裂应变等波动性下降,最终选择以规律性能较好的色差作为PC老化评估指标.通过Pearson相关性研究分析,认为各类环境因素与地理信息坐标等信息之间有信息冗余,环境参量可以进一步优选.通过灰色关联度、回归分析,筛选出影响PC老化的4个敏感环境因子分别为:日照时间、海拔高度、平均相对湿度、降水时间;通过构建反向传播人工神经网络(BP-ANN)模型并优化模型参数,建立具有良好训练精度及泛化能力的“环境-材料”映射模型.不同学习精度训练结果表明,过低的学习精度将导致训练程度不够,预测精度较低;过高的学习精度将导致过拟合,使预测陷入局部最小值.输入西藏全域28个城市气象数据,输入已训练好的人工神经网络模型,预测得到西藏28个城市PC老化一年色差值.基于Griddate插值计算,得到西藏地区严酷度空间分布地图,结果显示低海拔的藏东严酷度较低,而高海拔藏西无人区及藏北高原等地区严酷度较高.
The spatiotemporal distribution of atmospheric environmental severity in Tibet was evaluated and predicted based on polycarbonate(PC)chromatic aberrations.This study collected the annual average data(April 2021 to March 2022)of eight types of environmental factors from 10 typical atmospheric sites in Tibet.The climatic characteristics and climate distribution areas were analyzed to obtain accurate input to evaluate environmental severity.Natural environmental tests were conducted at 10 sites to analyze the regulation of PC degradation.The results showed that the gloss decreased and chromatic aberration gradually increased during PC aging,and mechanical properties,such as tensile strength and tensile strain at break,decreased with fluctuations.Thus,chromatic aberration was selected as a PC aging evaluation index owing to its excellent performance.Pearson’s correlation analysis was used to determine the information redundancy hidden in various environmental factors and geographic information coordinates.The environmental parameters were further optimized,and the factors highly related to PC aging were sunshine time,altitude,average relative humidity,and precipitation time.The“environmental material”mapping model with excellent training precision and genera-lizability was established using the Back Propagation Artificial Neural Network.By inputting the environmental data of 28 cities in Tibet into the well-built models,the severity was predicted and visualized to form spatial distribution maps using the Griddate interpolation method.The results showed that the low-altitude areas in eastern Tibet presented low severity.By training with different learning accuracies,the results revealed that low learning precision caused insufficient training and led to low prediction accuracy,whereas high learning precision led to overfitting and a prediction of the local minimum.The meteorological data of the 28 cities in Tibet were loaded into a well-trained artificial neural network model to predict the chromatic aberration
作者
吴德权
覃粒
谭甜甜
陈星昊
赵方超
张达威
杜翠薇
WU Dequan;QIN Li;TAN Tiantian;CHEN Xinghao;ZHAO Fangchao;ZHANG Dawei;DU Cuiwei(Southwest Institute of Technology and Engineering,Chongqing 400039,China;orrosion and Protection Center,University of Science and Technology Beijing,Beijing 100083,China)
出处
《工程科学学报》
EI
CSCD
北大核心
2024年第5期863-874,共12页
Chinese Journal of Engineering
基金
重庆市博士后基金资助项目(cstc2021jcyj-bshX0039)。
关键词
聚碳酸酯
西藏大气环境
人工神经网络
严酷度
预测
polycarbonate(PC)
atmospheric environment of Tibet
artificial neural network
severity
prediction
