基于BP神经网络的花椒干燥品质快速预测模型研究

Study on Fast Prediction Model of Pepper Drying Quality Based on BP Neural Network

随着现代农业的发展,花椒干燥品质检测已成为高效、快速、精准监测花椒干燥质量的重要一环。从花椒热风干燥最优温度40℃条件出发,并通过实验分析得出色泽(明亮度、红绿值、黄蓝值)、总黄酮和多酚等花椒主要干燥品质与损耗系数D与湿基含水率具有相关性,故以此建立基于BP神经网络预测花椒干燥品质的模型,结果表明:各品质参数预测值与实测值的相关系数均在0.93以上,均方根误差均在0.08以下。同时进一步验证预测模型,预测值与实测值的最大误差都在5%以内。该模型能够较好地预测花椒干燥品质,且能满足工程精度的要求,为花椒干燥品质的无损检测自动化、信息化和智能化水平提供了数据和理论支持。

With the development of modern agriculture, pepper drying quality detection has become an important part of efficient, rapid and accurate monitoring of pepper drying quality. Based on the optimal temperature of hot air drying of pepper at 40 ℃, it is concluded that the main drying quality such as color(brightness, red green value, yellow blue value), total flavonoids and polyphenols are correlated with loss coefficient D and moisture content on wet basis. So a model based on BP neural network is established to predict the drying quality of pepper, the results show that the correlation coefficients between the predicted and measured values of each quality parameter are above 0.93, and the root mean square error is below 0.08. Meanwhile, the prediction model is further verified, and the maximum error between the predicted value and the measured value is within 5%. This prediction model can better predict the drying quality of pepper,and can meet the requirements of engineering accuracy. It provides data and theoretical support for the automation, informatization and intelligence level of nondestructive testing of pepper drying quality.