宫兴, 英红, 姜鹏.基于BP神经网络的沥青混合料级配检测研究[J].河南理工大学学报（自然科学版）,2022,41(3):165-171.

GONG X , YING H, JIANG P.Research on asphalt mixture grading detection based on BP neural network[J].Journal of Henan Polytechnic University(Natural Science) ,2022,41(3):165-171.

基于BP神经网络的沥青混合料级配检测研究

宫兴, 英红, 姜鹏

桂林电子科技大学 建筑与交通工程学院，广西 桂林541004

摘要:为了准确、简便检测沥青混合料的设计级配，基于图像处理，通过改进电子筛分方式，将集料颗粒的平面形状分为三类分别进行统计，得到沥青混合料的平面级配，然后使用BP神经网络对沥青混合料的设计级配进行检测，以不同粒径粗集料的平面级配作为输入层，以不同粒径粗集料的设计级配作为输出层，对200组归一化处理后的平面级配进行神经网络训练和测试。结果表明：改进后的电子筛分方式较直接使用等效直径法和等效椭圆短轴法检测沥青混合料级配具有更高的正确率；使用BP神经网络对设计级配进行检测，各粒径的平均正确率分别为 88.1% （4.75 mm） ，91.2% （9.5 mm） ，93.8% （13.2 mm） ，95.1% （16 mm） ，100% （19 mm）,平面级配与设计级配的相关性较好。该方法为沥青混合料的级配检测提供了一种新思路。

关键词:沥青混合料;图像处理;电子筛分;BP神经网络;级配检测

doi:10.16186/j.cnki.1673-9787.2020060076

基金项目:国家自然科学基金资助项目（51968011 ；51668012）；桂林电子科技大学研究生优秀学位论文培育项目（18YJPYSS35 ）；桂林电子科技大学研究生教育创新计划资助项目（2018YJCX90）

收稿日期:2020/06/28

修回日期:2020/10/24

出版日期:2022/05/15

Research on asphalt mixture grading detection based on BP neural network

GONG Xing , YING Hong , JIANG Peng

School of Architecture and Transportation Engineering，Guilin University of Electronic Technology， Guilin 541004，Guangxi，China

Abstract: To accurately and easily detect the design gradation of asphalt mixture , the electronic sieving method was improved based on image processing, and the planar shapes of the aggregate particles was divided into three categories to obtain statistics of the planar gradation of asphalt mixture. The BP neural network was used to detect the design gradation of asphalt mixture,the planar gradation of coarse aggregates with different particle sizes was used as the input layer,and the design gradation was used as the output layer,200 of normalized planar gradations were trained and tested for neural networks. The results showed that the improved electronic sieving method had higher accuracy than the direct equivalent diameter method and the equivalent ellipse short axis method. The average accuracy of each particle size after detecting the design gradation using BP neural network were:88.1%(4.75 mm),91.2%(9.5 mm),93.8%(13.2 mm),95.1%(16 mm),100%(19 mm),respectively, and the correlation between planar gradation and design gradation was good. The proposed method could provide a new idea for the detection of asphalt mixture gradation.

Key words:asphalt mixture;image processing;electronic sieving;BP neural network;gradation detecting

基于BP神经网络的沥青混合料级配检测研究_宫兴.pdf