文章摘要
马子媛,李海莲,蔺望东.基于PCA-IPSO-RBF神经网络的沥青路面破损状况预测[J].,2022,62(2):197-205
基于PCA-IPSO-RBF神经网络的沥青路面破损状况预测
Prediction of asphalt pavement damage condition based on PCA-IPSO-RBF neural network
  
DOI:10.7511/dllgxb202202012
中文关键词: 道路工程  路面破损状况预测  径向基神经网络  沥青路面  改进粒子群优化  主成分分析
英文关键词: road engineering  prediction of pavement damage condition  radical basic function neural network  asphalt pavement  improved particle swarm optimization  principal component analysis
基金项目:国家自然科学基金资助项目(51868042);甘肃省自然科学基金资助项目(20JR10RA229);甘肃省青年科学基金资助项目(17JR5RA087);甘肃省高等学校创新基金资助项目(2021A048);甘肃省高等学校创新能力提升项目(2019B055);兰州交通大学青年科学基金资助项目(2017016);兰州交通大学“百名青年优秀人才培养计划”基金资助项目(2018103).
作者单位
马子媛,李海莲,蔺望东  
摘要点击次数: 64
全文下载次数: 47
中文摘要:
      针对现有路面破损状况评价指数较为单一且无法准确预测路面实际破损发展趋势的问题,为探究路面使用指标、路面性能指标和路面环境指标对路面破损状况的影响,提出基于改进粒子群优化(IPSO)与径向基(RBF)神经网络耦合的沥青路面破损预测模型.首先,利用灰色关联分析和主成分分析(PCA)筛选出主要影响因子;然后,通过改进粒子群惯性权重因子,调整粒子全局和局部寻优能力,并利用IPSO算法训练RBF模型中的参数;最后,以主成分分析降维后数据为输入,建立路面破损状况的IPSO-RBF神经网络预测模型.实例研究表明,PCA-IPSO-RBF神经网络预测模型预测平均绝对误差为0.841 6.因此,针对复杂非线性路面破损状况预测问题,该模型能够准确预测沥青路面破损状况,为路面养护决策提供有力支持.
英文摘要:
      In view of that the existing pavement damage condition evaluation is a single index and can not predict the development trend of the road actual damage, a damage prediction model is proposed for asphalt pavement which couples improved particle swarm optimization (IPSO) and radical basic function (RBF) neural network to explore the influence of pavement usage index, pavement performance index and road environment index on pavement damage condition. Firstly, the main influencing factors are screened out by grey correlation analysis and principal component analysis (PCA). Then, the global and local optimization abilities of particles are adjusted by improving the particle swarm inertia weight factor, and the IPSO algorithm is used to train the parameters in the RBF model. Finally, the IPSO-RBF neural network prediction model of pavement damage condition is established by taking the dimension reduction data of principal component analysis as input. Case study shows that the average absolute error of PCA-IPSO-RBF neural network prediction model is 0.841 6. Therefore, this model can accurately predict the damage of asphalt pavement and provide strong support for pavement maintenance decision-making.
查看全文   查看/发表评论  下载PDF阅读器
关闭