基于优化SSA-BP模型的牛肉微波干燥水分含量与能耗预测研究

凌菁¹，徐洁^2*，Dennis R. Heldman³，吴庭¹

(1. 广州航海学院计算机学院，广州510725，中国；

2. 华盛顿州立大学生物系统工程系，普尔曼WA 99164，中国；

3. 俄亥俄州立大学食品科学与技术系，哥伦布 OH 43210，中国)

摘要：该研究旨在探究一种增强型反向传播（Back-Propagation, BP）神经网络模型在牛肉微波干燥过程分析与预测中的应用价值。基于菲克第二扩散定律（Fick’s second law），实验测定了在不同微波功率（70-420 W）及相对湿度（0%、30%、50%）条件下的有效水分扩散系数。结果表明，有效水分扩散系数的变化范围为2.23×10^-9至2.87×10^-8 ㎡/s。研究发现，微波功率与单位能耗之间呈显著负相关关系；当微波功率为 420 W 时，系统能效达到最优（最低单位能耗为 8.39 MJ/kg 水）。

为精确模拟干燥动力学及能耗特征，本研究构建了多层BP神经网络架构，并引入麻雀搜索算法（Sparrow Search Algorithm, SSA）对网络的权重和阈值参数进行校准优化。对比分析显示，SSA优化的BP神经网络在预测精度方面显著优于传统BP模型及遗传算法（GA）优化的同类模型。该改进模型在预测干燥全过程的水分含量演变及能耗动态方面表现出优异的鲁棒性。

上述研究成果为开发兼顾产品品质与高能效的工业化牛肉干燥系统提供了重要的理论依据与技术支撑。该研究所提出的智能计算框架可为食品加工领域微波干燥过程的精准建模、预测及优化提供一种极具潜力的技术路径。

论文研究的意义和价值：牛肉微波干燥是一个涉及电磁场、热场与湿场耦合的高度非线性时变过程。传统机理模型在处理多变量干扰时往往存在偏差。本研究提出的SSA-BP混合模型，利用麻雀搜索算法极强的全局寻优能力，克服了传统BP神经网络收敛速度慢、易陷入局部最优的缺陷，实现了对干燥动力学参数的高精度逼近。这为复杂食品加工过程的数字化建模提供了新的算法范式。将该智能计算框架拓展至其他热敏性农产品（如高附加值果蔬、药材等）的微波干燥研究中，建立不同物料的介电特性与干燥动力学数据库，验证SSA-BP算法的泛化能力，将有助于形成通用的食品微波干燥智能预测平台。

关键词：微波功率；SSA-BP；单位能耗；水分含量；预测

DOI: 10.25165/j.ijabe.20251804.9832

引用信息: Ling J, Xu J, Heldman D R, Wu T. Prediction of moisture content and energy consumption in microwave drying of beef based on an optimized SSA-BP model. Int J Agric & Biol Eng, 2025; 18(4): 312–320.

点击阅读全文：Prediction of moisture content and energy consumption in microwave drying of beef based on an optimized SSA-BP model

点击合集阅读：AFPS（农产品和食品加工系统）

Prediction of moisture content and energy consumption in microwave drying of beef based on an optimized SSA-BP model

Jing Ling¹, Jie Xu^2*, Dennis R. Heldman³, Ting Wu¹

(1. School of Computer Science, Guangzhou Maritime University, Guangzhou 510725, China;

2. Department of Biological Systems Engineering, Washington State University, Pullman, WA 99164, USA;

3. Department of Food Science and Technology, The Ohio State University, Columbus, OH 43210, USA)

Abstract: This study investigates the application of an enhanced Back-Propagation (BP) neural network model for analyzing and predicting beef microwave drying processes. Based on Fick’s second law of diffusion, effective moisture diffusivity was determined  under varying  microwave  power  levels  (70-420  W)  and  relative  humidity conditions (0%, 30%, 50%). Experimental results revealed moisture diffusivity values ranging from  2.23×10^–9  to  2.87×10^–8 m²/s.  A  significant  inverse relationship was  observed between microwave power and specific energy consumption, with optimal energy efficiency (8.39 MJ/kg water) achieved at 420 W. A multi-layer BP neural network architecture was developed to model drying kinetics and energy consumption patterns, with subsequent optimization using Sparrow Search Algorithm (SSA) for weight and threshold parameter calibration. Comparative analysis demonstrated that the SSA-optimized BP neural network significantly outperformed both conventional BP models and genetic algorithm-optimized variants in predictive accuracy. The enhanced model exhibited robust performance in predicting moisture content evolution and energy consumption dynamics throughout the drying process. These findings provide valuable insights for developing energy-efficient industrial-scale beef drying systems while maintaining product quality. The proposed intelligent computing framework represents a promising approach for precise modeling, prediction, and optimization of microwave drying processes in food processing applications.

Keywords: microwave power, SSA-BP, specific energy consumption, moisture content, prediction

DOI: 10.25165/j.ijabe.20251804.9832

Citation: Ling J, Xu J, Heldman D R, Wu T. Prediction of moisture content and energy consumption in microwave drying of beef based on an optimized SSA-BP model. Int J Agric & Biol Eng, 2025; 18(4): 312–320.

往期推荐：

中国农业大学董仁杰教授等：可再生能源技术跨境转移对全球减排的贡献：中国-埃塞俄比亚光伏和沼气项目案例研究

华南农业大学彭红星副教授等：基于改进YOLOv8-pose模型的荔枝采摘机器人果实与采摘点同步检测方法

山东农业大学刘平教授等：基于RealSense传感器的小麦植株三维重建及表型解析

四川大学赵龙副教授等：基于无人机遥感数据估算苹果园土壤水分含量：LST/LAI两阶段特征空间理论的应用

中国农业大学王新副教授等：基于改进YOLOv5s的复杂环境下轻量级烟叶检测方法

支持单位：