A Design Method for an SVM-Based Humidity Sensor for Grain Storage

• Published in: Sensors (Basel, Switzerland) Vol. 24; no. 9; p. 2854
• Main Authors: Liu, Lining, Song, Chengbao, Zhu, Ke, Liu, Pingzeng
• Format: Journal Article
• Language: English
• Published: Switzerland MDPI AG 01.05.2024
• Subjects: Accuracy; Agricultural machinery; Agriculture; Backup software; Calibration; Costs; Crop yields; Farm equipment; Farm produce; Food security; Food supply; Grain; grain moisture; Humidity; humidity sensor; Methods; Moisture absorption; moisture calibration models; Mold; Moving & storage industry; Rice; Safety and security measures; Sensors; standing wave ratio (SWR) method; Ventilation; China; standing wave ratio (SWR) method; humidity sensor; moisture calibration models; grain moisture
• DOI: 10.3390/s24092854
• ISSN: 1424-8220

One of the crucial factors in grain storage is appropriate moisture content, which plays a significant role in reducing storage losses and ensuring quality. However, currently available humidity sensors on the market fail to meet the demands of modern large-scale grain storage in China in terms of price, size, and ease of implementation. Therefore, this study aims to develop an economical, efficient, and easily deployable grain humidity sensor suitable for large-scale grain storage environments. Simultaneously, it constructs humidity calibration models applicable to three major grain crops: millet, rice, and wheat. Starting with the probe structure, this study analyzes the ideal probe structure for grain humidity sensors. Experimental validations are conducted using millet, rice, and wheat as experimental subjects to verify the accuracy of the sensor and humidity calibration models. The experimental results indicate that the optimal length of the probe under ideal conditions is 0.67 m. Humidity calibration models for millet, rice, and wheat are constructed using SVM models, with all three models achieving a correlation coefficient R2 greater than 0.9. The measured data and model-calculated data show a linear relationship, closely approximating y = x, with R2 values of all three fitted models above 0.9. In conclusion, this study provides reliable sensor technological support for humidity monitoring in large-scale grain storage and processing, with extensive applications in grain storage and grain safety management.