摘要
利用自制粮食孔隙率测定仪,采用直接测量法对不同受压状态下的粮食单元体孔隙率进行测量,得到不同粮种、不同含水率和不同压力下的粮食单元体孔隙率。通过粒子群算法(PSO)优化支持向量回归(SVR),建立基于PSO-SVR粮食单元体孔隙率的预测模型,并与随机森林(RF)模型、SVR模型对比分析其性能。结果表明:PSO-SVR模型的各项性能指标均优于RF模型和SVR模型。PSO-SVR模型测试样本的均方误差(MSE)为0.0660、决定系数(R2)为0.9340、平均绝对误差(MAE)为0.2000,相较其他2种模型,该模型的预测结果误差小,具有较高的预测精度,可以有效预测粮食在不同压力下的孔隙率。
The direct measurement method was used to measure the porosity of grain units under different pressure by self-made grain porosity tester.The porosity of grain units under different grain varieties,different water content and different pressure was obtained.Particle swarm optimization(PSO)was used to optimize support vector regression(SVR),and a prediction model of grain units porosity based on PSO-SVR was established,and its performance was compared with random forest(RF)model and SVR model.The results showed that PSO-SVR model was superior to RF model and SVR model in all performance indexes.The mean square error(MSE),determination coefficient(R°)and mean absolute error(MAE)of the PS0-SVR model were 0.0660,0.9340 and 0.2000,respectively.Compared with the other two models,the prediction error of PSO-SVR model is small,and it has high prediction accuracy,which can effectively predict the porosity of grain under different pressures.
作者
陈家豪
郑倩茹
金立兵
郑德乾
尹君
李嘉欣
CHEN Jia-hao;ZHENG Qian-ru;JIN Li-bing;ZHENG De-qian;YIN Jun;LI Jia-xin(College of Civil Engineering,Henan University of Technology,Zhengzhou 450001,Henan,China;Henan Modern Green Ecological Storage System International Joint Laboratory,Zhengzhou 450001,Henan,China;Academy of National Food and Strategic Reserves Administration,Beijing 100037,China)
出处
《粮食与油脂》
北大核心
2024年第6期55-59,共5页
Cereals & Oils
基金
国家自然科学基金(51608176)
河南省科技研发计划联合基金(应用攻关类)(222103810082、232103810080)
河南工业大学青年骨干教师培育计划(21420155)。
关键词
粮食
孔隙率
机器学习
粒子群算法
支持向量回归
grain
porosity
machine learning
particle swarm optimization
support vector regression