论文部分内容阅读
目的:建立人工神经网络用于预测纤维性根茎药材的超滤总多糖保留率。方法:以无机陶瓷膜的膜孔径、滤过压力、药液温度为输入变量,不同超滤条件下红芪酶解提取液中总多糖保留率为输出变量,采用Levenberg-Marquardt算法优化网络参数,建立BP神经网络预测模型,并对模型的性能及适用性进行考察。结果:BP神经网络的拓扑结构为3-6-1,对红芪总多糖保留率预测的平均预测误差、平均绝对误差和平均误差率分别为0.10%,0.98%,1.55%;对黄芪总多糖保留率预测的平均误差率2.77%。结论:建立的模型预测精度较高,适用性较好,可用于预测纤维性根茎药材超滤的总多糖保留率。
Objective: To establish an artificial neural network for predicting the total polysaccharide retention rate of fibrous rhizomes. Methods: The membrane pore size, filtration pressure and liquid temperature of inorganic ceramic membranes were used as input variables. The total polysaccharide retention rate of extracts from Radix Astragali was determined as output variables under different ultrafiltration conditions. The network parameters were optimized by Levenberg-Marquardt algorithm. BP neural network prediction model is established, and the performance and applicability of the model are investigated. Results: The topological structure of BP neural network was 3-6-1. The average prediction error, the average absolute error and the average error rate of prediction of the total polysaccharide retention rate of Radix Astragali were 0.10%, 0.98% and 1.55% The average rate of error of the retention rate forecast is 2.77%. Conclusion: The established model has higher prediction accuracy and better applicability, which can be used to predict the total polysaccharide retention rate of fibrous rhizomes by ultrafiltration.