Authors

Abstract

Addressing the issue of high cleaning loss rates encountered during actual combine harvester operations, this study designed a detection device specifically for monitoring cleaning losses. Initially, a three-dimensional model of wheat grains was established using Blender software. Subsequently, the impact processes of wheat grains and straw falling from different heights onto a sensitive plate were simulated using EDEM discrete element analysis software, from which contact force variation curves and motion trajectories were obtained. The results indicated a significant difference in the impact forces of the two material types on the sensitive plate, enabling material identification and loss rate calculation through signal acquisition. Based on these findings, a detection device comprising a mechanical structure and a control system was developed. An ESP32 microcontroller was employed to read data from piezoelectric ceramic vibration sensors. After processing the data with a Kalman filter, material classification thresholds were determined based on the principles of normal distribution. Preliminary experimental parameters were established through a three-factor, three-level experiment, and subsequently optimized using response surface methodology. The experimental results demonstrated that optimal threshold differentiation and the highest accuracy in loss rate calculation were achieved under the following conditions: a sensitive plate installation height of 550 mm, an inclination angle of 40°, and a conveyor belt speed of 8 m/min. Bench tests verified that the overall error of the device was less than 3%, with recognition rates exceeding 97% for both wheat grains and straw.

Abstract in Chinese

针对实际作业中清选损失率偏高的问题，本研究设计了一种用于清选损失检测的装置。先利用Blender软件建立小麦籽粒的三维模型，再通过EDEM离散元分析软件模拟不同高度下落的小麦籽粒与秸秆对敏感板的撞击过程，获取接触力变化曲线和运动轨迹。结果显示，两类物料撞击敏感板的力度差异显著，可通过信号采集实现物料识别并计算损失率。在此基础上，设计了包含机械结构和控制系统的检测装置，采用ESP32单片机读取压电陶瓷振动传感器数据，经卡尔曼滤波处理后，结合正态分布规律确定物料分类阈值。通过三因素三水平实验初步确定实验参数并通过响应曲面确定实验参数，实验结果表明，当敏感板的安装高度为550mm、倾斜角度为40°、传送带速度8米/分钟时，阈值区分度最佳，损失率计算精度最高。台架实验验证表明，该装置整体误差小于3%，对小麦籽粒和秸秆的识别率均达97%以上。