基于PSO-ANN的再生-天然基层填料回弹模量预测  

Prediction for resilient modulus of recycled aggregates-natural aggregate base filler based on PSO-ANN

作　　者：肖源杰 王政 李昀博 王萌 孔坤锋 周震 XIAO Yuanjie;WANG Zheng;LI Yunbo;WANG Meng;KONG Kunfeng;ZHOU Zhen(School of Civil Engineering,Central South University,Changsha,Hunan 410075,China;MOE Key Laboratory of Engineering Structure of Heavy Haul Railway(Central South University),Changsha,Hunan 410075,China;Railway Engineering Research Institute,China Academy of Railway Sciences Corporation Limited,Beijing 100081,China;Guangdong Communication Planning and Design Institute Group Co.Ltd.,Guangzhou,Guangdong 510440,China)

机构地区：[1]中南大学土木工程学院,湖南长沙410075 [2]重载铁路工程结构教育部重点实验室,湖南长沙410075 [3]中国铁道科学研究院集团有限公司铁道建筑研究所,北京1000811 [4]广东省交通规划设计院集团股份有限公司,广东广州510440

出　　处：《河北工业大学学报》2025年第1期79-86,共8页Journal of Hebei University of Technology

基　　金：湖南省自然科学基金杰出青年基金资助项目(2024JJ2073);国家自然科学基金资助项目(52178443);交通运输部重点科技项目(2022-MS5-122);国家重点研发计划资助项目(2019YFC1904704);中南大学中央高校基本科研业务费专项资金资助项目(2023ZZTS0019)。

摘　　要：为探究建筑固废再生骨料在道路工程中大规模和高质量应用的可行性,针对再生骨料-天然骨料二元混合试样,开展了一系列不同再生骨料取代率、含水率、压实度以及围压和动偏应力组合条件下的室内大型动三轴试验;建立了粒子群优化人工神经网络回弹模量预测模型,并采用蒙特卡洛方法对比分析了本文模型与其他几种目前常见的机器学习模型的鲁棒性和泛化能力。研究结果表明:动回弹模量各影响因素按相关性和显著性水平排序为围压>动偏应力幅值>含水率>压实度>再生骨料取代率,动回弹模量随围压变化近似呈线性增长,含水率的降低或压实度的提高均会导致动回弹模量增大,不同再生骨料取代率试样的动回弹模量差异不明显;相比于支持向量机和随机森林算法,粒子群优化人工神经网络模型具有较高的预测精度和鲁棒性。研究结果可为建筑固废再生基层填料的填筑以及相似颗粒材料的回弹特性评定提供参考。To investigate the feasibility of large-scale and high-quality application of recycled aggregates from construction and demolition waste in road engineering,a series of dynamic triaxial tests were conducted on binary mixed samples of recycled aggregate and natural aggregate under various conditions of recycled aggregate replacement rate,moisture content,compaction degree,and combinations of confining pressure and dynamic deviator stress.An artificial neural network model for predicting resilient modulus optimized by particle swarm optimization method was established.And the robustness of the model was comparatively analyzed with several other common machine learning models using the Monte Carlo method.The experimental research results show that,according to the level of relevance and significance,the order of various influencing factors on resilient modulus is confining pressure>dynamic deviator stress amplitude>moisture content>compaction degree>recycled aggregate replacement rate.The resilient modulus increases approximately linearly with the variation of confining pressure,and gradually increases with the decrease of moisture content or the increase of compaction degree.Compared with support vector machine and random forest algorithm,the particle swarm optimization artificial neural network model has higher prediction accuracy and robustness.The research results can provide a reference for the construction of recycled aggregates filling and the evaluation of the resilient characteristics of similar granular materials.

关 键 词：建筑固废 再生骨料 回弹模量 预测模型 机器学习 

分 类 号：U414[交通运输工程—道路与铁道工程]