摘要
在不同单位辐射功率和辐射距离下对山药切片进行了红外辐射干燥温度试验。基于温度试验数据,通过Matlab神经网络工具箱,采用最速梯度下降法和L-M算法对这些数据分别进行了训练,将训练好的BP神经网络对山药切片进行温度预测。结果表明:L-M算法优于传统的最速梯度下降法,提高了BP神经网络的收敛速度和泛化能力,预测误差较小,适用性较强,可较好地预测红外干燥过程中山药切片的温度变化。
Infrared drying experiments were carried out and the temperature data of yam were collected under different infrared intensities and infrared distances. The experiment results showed that the infrared intensity, infrared distance and drying time played an important role on the surface temperature and internal temperature of yam. Thus, infrared intensity, infrared distance and drying time were chosen as the input layers vectors of BP neural network model. A 3 × 9 × 1 single hidden layer BP network model was established. The model was trained by steepest gradient descent method and Levenberg- Marquardt algorithm respectively based on temperature data of yam. The maximum prediction error of optimized network model using Levenberg- Marquardt algorithm was 1.3℃ , while the traditional algorithm of BP neural network was 5.7 ℃. It was indicated that Levenberg - Marquardt optimization method was superior to the steepest gradient descent method in the predicting temperature of yam with high precision. Therefore, it is feasible to predict temperature variations of materials during infrared drying process by using BP neural network model optimized by L- M algorithm.
出处
《农业机械学报》
EI
CAS
CSCD
北大核心
2014年第11期246-249,336,共5页
Transactions of the Chinese Society for Agricultural Machinery
基金
国家自然科学基金资助项目(31271908)
山东省农业科技成果转化资助项目(鲁科农字[2012]65号)
关键词
山药
红外干燥
温度
L—M算法
神经网络
Yam Infrared drying
Temperature
Levenberg- Marquardt algorithm
Neural network
