|国家科技期刊平台
首页|期刊导航|农机化研究|基于机器学习的农田土壤抗剪强度参数检测方法研究

基于机器学习的农田土壤抗剪强度参数检测方法研究OA北大核心

Research on the Detection Method of Farmland Soil Shear Strength Parameters Based on Machine Learning

中文摘要英文摘要

土壤抗剪强度参数包括粘聚力和内摩擦角,是评价土壤侵蚀敏感性和反映耕层耕作性能的重要指标.为实现农田土壤抗剪切强度参数的快速检测,提出了一种基于机器学习的土壤抗剪切强度参数检测方法.以STM32 单片机为核心处理器,采用圆锥杆、滚珠丝杆滑台、三角支架等构建土壤数据采集装置,利用DYMH-103 柱式压力传感器和FlexiForce薄膜传感器分别检测圆锥杆贯入土壤的锥尖阻力和锥侧压力,采用CSF11 土壤水分传感器获取土壤含水率信息,通过多传感器数据特征向量提取构建建模数据集.数据集相关性分析结果表明:土壤抗剪强度参数与锥尖阻力、锥侧压力和土壤含水率之间具有明显相关性.利用蒙特卡罗交叉验证(Monte Carlo Cross Validation,MCCV)剔除了数据集中的 4 个异常样本;同时,提出了一种ELM-PLSR组合建模算法,以决定系数R2 和RPD为评价指标,对比评估了ELM、PLSR和ELM-PLSR 3 种不同机器学习模型,结果表明:ELM-PLSR模型预测性能优于ELM模型和PLSR模型;检测粘聚力时,对应的R2、RPD分别为 0.919 和 3.475;检测内摩擦角时,对应的R2 和RPD分别为 0.910 和 3.304.研究结果可为土壤抗剪强度参数快速测量提供参考.

Soil shear strength parameters,including cohesion and internal friction angle,are important indexes to evalu-ate soil erosion sensitivity and reflect tillage performance in topsoil.In order to realize the rapid detection of soil shear strength parameters,a method of soil shear strength parameters detection based on machine learning was proposed.With an STM32 micro-controller as the core processor,the soil data acquisition device was constructed with cone rod,ball screw sliding table and triangle bracket.A DYMH-103 column pressure sensor and a FlexiForce film sensor were used to detect cone tip resistance and cone lateral pressure of cone rod penetration into soil respectively,and a CSF11 soil mois-ture sensor was used to obtain soil moisture content information.The feature vectors were extracted from the data collected by these sensors to construct the modeling data set.Then the Monte Carlo Cross Validation(MCCV)was used to elimi-nate 4 abnormal samples from the dataset.Furthermore,an ELM-PLSR combined modeling algorithm was proposed.Finally,three different machine learning models,ELM,PLSR and ELM-PLSR,were compared and evaluated using the determination coefficient R2 and RPD as evaluation indexes.The results show that the predictive performance of the ELM-PLSR model is better than that of the ELM model and the PLSR model.The R2 and RPD of ELM-PLSR to the co-hesion detection are 0.919 and 3.475,and to the internal friction angle detection are 0.910 and 3.304,respectively.This study can provide a reference method for the rapid measurement of soil shear strength parameters.

于艳艳;朱龙图;刘鹤

吉林农业大学 信息技术学院,长春 130118云南农业大学 机电工程学院,昆明 650201||农业农村部长江中下游农业装备重点实验室,武汉 430070

农业科学

机器学习土壤抗剪强度多传感器特征向量预测模型

machine learningsoil shear strengthmultiple sensorfeature vectorprediction model

《农机化研究》 2025 (001)

7-15 / 9

吉林省科技厅中青年科技创新创业卓越人才(团队)项目(20220508133RC);中国博士后科学基金项目(2021M701341)

10.13427/j.issn.1003-188X.2025.01.002

评论