摘要
筒仓卸料时的动态侧压力是导致筒仓结构被破坏的重要原因,但影响筒仓动态侧压力的因素众多,且相互之间存在着复杂的非线性关系。因此,建立一种考虑多因素影响、高效、准确的动态侧压力的预测方法尤为重要。基于机器学习方法,将支持向量机、BP神经网络和随机森林等3种机器学习方法应用到筒仓动态侧压力的预测中。选取影响筒仓动态侧压力的相关因素作为输入变量,动态侧压力为输出值。对常用的3种机器学习中的参数进行寻优与设置,建立筒仓动态侧压力预测模型。通过测试样本对预测模型进行测试,分析表明支持向量机算法具有最优的预测能力与适用性,为筒仓动态侧压力的预测提供了一种新的方法。通过MATLAB软件对贮料密度这一单因素进行随机抽样,得到1 000组均匀分布的随机数。将数据输入最优的预测模型中,并利用Easyfit软件对预测值进行概率分布拟合,得到筒仓动态侧压力的概率分布,为筒仓结构的可靠度研究提供了理论基础。
Dynamic lateral pressure during silo unloading is an important cause of silo structure damage.However,there are many factors affecting the dynamic lateral pressure of silo,and there is a negative complex nonlinear relationship among them.Therefore,it is particularly important to establish an efficient and accurate prediction method for dynamic lateral pressure considering multiple factors.Based on the machine learning method,three machine learning methods of support vector machine,BP neural network and random forest were applied to predict the dynamic lateral pressure of silo.Firstly,the relevant factors affecting the dynamic side pressure of silo were selected as input variables,and the dynamic side pressure was the output value.Then,the parameters of three commonly used machine learning were optimized and set,and the dynamic lateral pressure prediction model of silo was established.The prediction model was tested by test samples,and the analysis showed that the support vector machine algorithm had the optimal prediction ability and applicability,which provided a new method for the prediction of dynamic lateral pressure of silo.Finally,MATLAB software was used to conduct random sampling on the single influencing factor of storage density,and 1 000 groups of uniformly distributed random numbers were obtained.The data was input into the optimal prediction model,and the probability distribution of the predicted value was fitted by Easyfit software.The probability distribution of the dynamic lateral pressure of the silo was obtained,which provided a theoretical basis for the reliability study of the silo structure.
作者
余汉华
徐志军
赵世鹏
刘婷婷
原方
YU Hanhua;XU Zhijun;ZHAO Shipeng;LIU Tingting;YUAN Fang(College of Civil Engineering,Henan University of Technology,Zhengzhou 450001,China)
出处
《河南工业大学学报（自然科学版）》
CAS
北大核心
2022年第2期103-110,共8页
Journal of Henan University of Technology：Natural Science Edition
基金
国家自然科学基金项目(51578216)
河南省高等学校青年骨干教师培养计划(2021GGJS058)。
关键词
动态侧压力
机器学习
预测模型
随机抽样
概率分布
dynamic lateral pressure
machine learning
prediction model
random sampling
probability distribution
