基于IFA-SVM的高速公路沥青路面使用性能预测  

Prediction of Performance of Expressway Asphalt Pavement Based on IFA-SVM

作　　者：李海莲[1] 林梦凯[1] 王起才[1] LI Hai-lian;LIN Meng-kai;WANG Qi-cai(School of Civil Engineering,Lanzhou Jiaotong University,Lanzhou Gansu 730070,China)

机构地区：[1]兰州交通大学土木工程学院

出　　处：《公路交通科技》2019年第12期8-14,78,共8页Journal of Highway and Transportation Research and Development

基　　金：国家自然科学基金项目(51868042);教育部长江学者和创新团队发展计划项目(IRT_15R29);甘肃省高等学校创新能力提升项目(2019B-055);甘肃省青年科学基金项目(17JR5RA087);兰州交通大学青年科学基金项目(2017016);兰州交通大学“百名青年优秀人才培养计划”基金项目(2018103)

摘　　要：针对传统定性法对高速公路沥青路面使用性能预测精度不高的问题,结合支持向量机理论和改进萤火虫算法,建立了一种基于IFA-SVM的预测模型。首先在预测模型中引入萤火虫领域搜索,克服了寻优过程中随着迭代次数的增加而发生萤火虫的随机移动。其次,在后续寻优过程中采用动态调整算法搜索步长来平衡全局搜索能力,加快了SVM模型性能参数的寻优选择。最后通过实例验证,并与标准FA-SVM预测方法进行对比分析,验证了IFA-SVM模型的有效性和预测精度的可行性。研究结果表明:(1)采用标准FA-SVM对G6高速公路白银段路面使用性能各个指标进行预测,其相对误差最大达2.543 5%,最小为0.820 6%,而利用IFA-SVM模型预测结果的相对误差最值分别为1.085 8%和0.365 4%,且其均方根误差均小于标准FA-SVM方法。(2)IFA-SVM模型在高速公路沥青路面使用性能预测时,收敛速度更快,精度高于标准的FA-SVM,预测结果不仅更加接近实测值,而且对高速公路沥青路面的养护决策提供有效支持。Aiming at the problem of the low accuracy of the traditional qualitative method for expressway asphalt pavement performance, a prediction model based on IFA-SVM is established by combining SVM theory and improved FA. First, the firefly field search is introduced into the prediction model to overcome the random movement of fireflies with the increase of the number of iterations in the optimization process. Second, in the subsequent optimization process, the step size is searched by using dynamic adjustment algorithm to balance the global search ability, which accelerates the optimization selection of the performance parameters of the SVM model. Finally, the example is illustrated and compared with the standard FA-SVM prediction method to verify the effectiveness of the IFA-SVM model and the feasibility of the prediction accuracy. The result shows that(1) The maximum relative error is 2.543 5% and the minimum is 0.820 6% when the standard FA-SVM is used to predict each indicator of the pavement performance of the Baiyin section of G6 expressway, while by using IFA-SVM model, the predicted maximum relative error is 1.085 8% and the minimum is 0.365 4%,and their root mean square errors are smaller than the standard FA-SVM method.(2) IFA-SVM model has a faster convergence rate and a higher accuracy than the standard FA-SVM when predicting the performance of asphalt pavement of expressway. The prediction result is not only closer to the measured value, but also provides an effective support for the maintenance decision of asphalt pavement of expressway.

关 键 词：道路工程 路面性能预测 领域搜索 高速公路 支持向量机 萤火虫算法 

分 类 号：U416.217[交通运输工程—道路与铁道工程]