河南农业科学 ›› 2021, Vol. 50 ›› Issue (9): 163-171.DOI: 10.15933/j.cnki.1004-3268.2021.09.020

• 农业信息与工程·农产品加工 • 上一篇    下一篇

基于BP 神经网络的畜禽粒状饲料水分测定系统设计

张洗玉1,朱果露2,陈雨欣2,高浩源2,李嘉乐2   

  1. (1.广西科技师范学院数学与计算机科学学院,广西来宾546199;2.陕西科技大学镐京学院,陕西西安712046)
  • 收稿日期:2021-01-10 出版日期:2021-09-15 发布日期:2021-10-13
  • 作者简介:张洗玉(1988-),男,陕西西安人,讲师,硕士,主要从事自动化控制技术及应用研究。E-mail:1102086190@qq.com
  • 基金资助:
    广西高校中青年教师科研基础能力提升项目(2021KY0857)

Design of Water Content Measurement System for Animal Granular Feed Based on BP Neural Network

ZHANG Xiyu1,ZHU Guolu2,CHEN Yuxin2,GAO Haoyuan2,LI Jiale2   

  1. (1.School of Mathematics and Computer Science,Guangxi Normal University of Science and Technology,Laibin 546199,China;2.Haojing College of Shaanxi University of Science and Technology,Xi’an 712046,China)
  • Received:2021-01-10 Published:2021-09-15 Online:2021-10-13

摘要: 为了解决传统畜禽粒状饲料水分含量测定过程中存在的成本耗费大、测定效率低、无法实现快速在线检测等问题,设计了间接式新型水分含量测定系统,用于畜禽粒状饲料在线快速测量。根据GB/T6435—2014要求及测定系统总体结构,以嵌入式芯片STM32F103C8T6为主控核心,选取差频式圆柱形电容传感器检测粒状饲料样品电容,使用温度传感器DS18B20测定样品测试环境温度,基于BP神经网络融合算法实现多传感器数据融合以补偿温度、紧实度等因素所造成的测定误差,根据多元回归分析方法建立粒状饲料样品水分测定模型,并利用MATLAB曲线拟合。结果表明,设计的测定系统能够在5~8 s内测定粒状饲料水分含量,测定绝对误差低于1.2%,测量重复性误差小于0.2。可见,设计的粒状饲料水分含量测定系统符合养殖业饲料生产精准管理且有效降低饲养成本的要求。

关键词: 畜禽粒状饲料, 水分含量, 测定系统, 无损在线检测, 水分模型, BP网络融合算法, MATLAB曲线拟合

Abstract: In order to solve the problems of high cost,low efficiency,and failing to realize rapid on⁃line detection in the process of traditional measurement of moisture content in the livestock and poultry granular feed,a new nondestructive indirect on⁃line rapid measurement system of moisture content was designed. According to the requirements of overall structure of measurement system and GB/T 6435—2014,the STM32F103C8T6 was used as microcontroller; the differential frequency cylindrical capacitance sensor was selected to detect the capacitance of the granular feed sample;the temperature sensor DS18B20 was used to measure the temperature of the sample test environment;the multi⁃sensor data fusion based on BP neural network algorithm compensated for the error by temperature,compactness and other factors.According to the method of multiple regression analysis,the moisture content model of granular feed sample was established and fitted with MATLAB curve.The experimental results showed that the system could determine the moisture content of granular feed in 5—8 s,the absolute error was less than 1.2%,and the repeatability error was less than 0.2.The system meets the new requirements of precise management of feed production and effective reduction of feeding cost.

Key words: Granular feed for livestock and poultry, Moisture content, Measurement system, Nondestructive online detection, Moisture model, BP network fusion algorithm, MATLAB curve fitting

中图分类号: