储粮仓仓壁动态侧压力的树模型预测方法  

Prediction Method of Dynamic Normal Stress on the Silo Wall Assisted with Tree Model

作　　者：徐志军[1,2] 彭舒停 赵世鹏 范量 余汉华 XU Zhi-jun;PENG Shu-ting;ZHAO Shi-peng;FAN Liang;YU Han-hua(School of Civil Engineering,Henan University of Technology,Zhengzhou 450001,China;Henan International Joint Laboratory of Modern Green Ecological Storage System,Zhengzhou 450001,China)

机构地区：[1]河南工业大学土木工程学院,郑州450001 [2]河南省现代绿色生态仓储体系国际联合实验室,郑州450001

出　　处：《科学技术与工程》2024年第26期11158-11166,共9页Science Technology and Engineering

基　　金：国家自然科学基金面上项目(51578216);河南省青年骨干教师培养项目(2021GGJS058)。

摘　　要：针对储粮仓卸料时仓壁动态侧压力难以准确预测的问题,利用机器学习方法中的树模型建立了仓壁动态侧压力预测模型。首先,分析了仓壁动态侧压力的主要影响因素为筒仓的结构尺寸、贮料的物理参数及测点位置。利用收集的496组仓壁动态侧压力数据,构建机器学习预测模型的数据集。然后,基于树模型,建立了仓壁动态侧压力的决策树(decision tree,DT)预测模型,在此基础上,利用Bagging算法和Boosting算法,建立了仓壁动态侧压力的随机森林(random forest,RF)预测模型和梯度提升树(gradient boosting decision tree,GBDT)预测模型。通过对比3种预测模型在测试集的均方误差(mean-square error,MSE)、决定系数和相对误差,表明GBDT预测模型的泛化性能最优。最后,通过开展模型试验和数值模拟,对GBDT预测模型进行验证,结果表明拟合良好。同时,根据树模型的分枝原理,判断出仓壁动态侧压力影响因素的重要性,得到对于贮料的物理参数,密度的重要性排第一;对于筒仓的结构尺寸,卸料口尺寸排第一。因此,在进行储粮仓设计时,建议优先考虑仓内散体物料的密度和仓的卸料口尺寸。It is difficult to accurately predict the normal stress on the silo wall during discharge.To address this issue,a prediction model is proposed for the dynamic normal stress on the silo wall,utilizing a tree model.Firstly,the main influencing factors of the dynamic normal stress on the silo wall were identified as the structural size of the silo,the physical parameters of the storage material,and the position of the measuring point.A data set for the machine learning prediction model was then constructed using the collected 496 groups of dynamic normal stress data on the silo wall.Subsequently,based on the tree model,a DT(decision tree) prediction model for the dynamic normal stress on the wall was established.On this foundation,the RF(random forest) prediction model for dynamic normal stress and the GBDT(gradient boosting decision tree) prediction model were established,employing the Bagging algorithm and the Boosting algorithm respectively.By comparing the MSE(mean square error),determination coefficient,and relative error of the three prediction models in the test set,it is shown that the GBDT prediction model exhibits the best generalization performance.Furthermore,the GBDT prediction model has been verified through model testing and numerical simulation,with satisfactory fitting results.Additionally,according to the branching principle of the tree model,the importance of the influencing factors for the dynamic lateral pressure of the silo is judged,indicating that the density of the storage material and the size of the discharge port rank first for the storage materials and silo structure respectively.Therefore,when designing a silo,it is recommended to prioritize the density of the bulk materials within the silo and the size of the discharge port.

关 键 词：储粮仓 动态侧压力 树模型 参数寻优 预测模型 

分 类 号：S379.9[农业科学—农产品加工]