In order to meet the increasing demand for food and beverage safety and quality, this study focused on the application of a back propagation (BP) neural network to determine the leaching rate of heavy metal in tea to improve the scientific health of tea drinking. The evaluation index and target expectations have been determined based on the extraction experiment of heavy metal Cd in tea soaking, with 3 evaluation index values taken as input layer parameters and the heavy metal extraction rate taken as output layer parameter. Then, employ the sample data standardized by min‐max linearization method to train and test the network model and get the satisfactory results, which showed that the constructed BP neural network expressed a fast convergence speed and the systematic error was as low as 0.0003509. Additionally, there was no significance between Cd leaching rate of experimental results and neural network model results by reliability testing with a correlation coefficient was .9895. These results revealed that the network model established possessed an outstanding training accuracy and generalization performance, which effectively reflected the extraction rate of heavy metal in tea soaking and improved the safety of tea drinking.

中文翻译：

---

基于反向传播神经网络的茶叶中重金属镉提取行为研究。

为了满足对食品和饮料安全性和质量日益增长的需求，本研究着重于应用反向传播（BP）神经网络来确定茶中重金属的浸出率，以改善茶饮的科学健康性。根据浸泡茶中重金属镉的提取实验，确定了评价指标和目标期望值，将3个评价指标值作为输入层参数，将重金属提取率作为输出层参数。然后，采用最小-最大线性化方法标准化的样本数据对网络模型进行训练和测试，并获得满意的结果，表明所构建的BP神经网络表示收敛速度快，系统误差低至0.0003509。另外，Cd浸出率与神经网络模型结果经可靠性测试无相关性，相关系数为.9895。这些结果表明，所建立的网络模型具有良好的训练精度和泛化性能，有效地反映了浸泡茶中重金属的提取率，提高了饮茶的安全性。