基于神经网络优化模型的中药复方安慰剂配色模拟研究  

Simulation Study on Color Matching of Traditional Chinese Medicine Compound Placebo Based on Neural Network Optimization Model

作　　者：李航[1] 黎盛强 周恩丽 王团结[2,3] 章晨峰 张欣[2,3] 肖伟 王振中[1,2,3] LI Hang;LI Shengqiang;ZHOU Enli;WANG Tuanjie;ZHANG Chenfeng;ZHANG Xin;XIAO Wei;WANG Zhenzhong(Kanion School of Chinese Materia Medica,Nanjing University of Chinese Medicine,Nanjing 210023,China;National Key Laboratory on Technologies for Chinese Medicine Pharmaceutical Process Control and Intelligent Manufacture,Lianyungang 222001,China;Jiangsu Kanion Pharmaceutical Co.,Ltd.,Lianyungang 222001,China)

机构地区：[1]南京中医药大学康缘中药学院,江苏南京210023 [2]中药制药过程控制与智能制造技术全国重点实验室,江苏连云港222001 [3]江苏康缘药业股份有限公司,江苏连云港222001

出　　处：《南京中医药大学学报》2024年第1期18-25,共8页Journal of Nanjing University of Traditional Chinese Medicine

基　　金：连云港市揭榜挂帅项目(CGJBGS2101)。

摘　　要：目的构建粒子群反向传播(Particle swarm optimization-back propagation,PSO-BP)神经网络对中药复方颗粒剂安慰剂制备着色剂的用量进行预测,为中药复方颗粒剂安慰剂颜色的模拟提供一种新思路。方法运用BP神经网络建立样品颜色参数L、a^(*)、b^(*)与色素质量分数的模型,利用粒子群算法的全局搜索能力优化BP神经网络权重和偏置,防止模型出现局部最小值,再采用线性降低权系数法和引入变异算子提高粒子群算法的全局寻优能力;以颜色综合评价指标(ΔE)为客观评价标准,验证试验结果。结果训练结果表明,改进的PSO-BP神经网络拟合精度最高达到98.31%;预测结果表明,改进的PSO-BP神经网络的预测误差最小,平均绝对百分比误差(MAPE)、均方根误差(RMSE)和平均色差(ΔE)分别为0.4115、2.1646、2.56;制备3种颗粒的验证样品进行验证,验证样品与模型药物的ΔE分别为1.73、2.63、4.11,肉眼直观评价其中两组与模型药物色差较小。结论基于改进粒子群优化算法的BP神经网络可模拟中药复方颗粒剂安慰剂制备着色剂用量预测,可作为安慰剂配色研究的推荐优化模型。OBJECTIVE To predict the amount of colorants used in the preparation of placebo of Chinese medicine by constructing particle swarm optimization-back propagation neural network PSO-BPNN compound granules,and to provide a new idea for the simulation of placebo color of Chinese medicine compound granules.METHODS The BP neural network was used to establish the model of sample color parameters L,a^(*),b^(*)and pigment mass fraction.The global search ability of particle swarm optimization algorithm was used to optimize the weight and bias of BP neural network to prevent the local minimum value of the model.The linear reduction weight coefficient method and the introduction of mutation operator were used to improve the global optimization ability of particle swarm optimization algorithm.The color comprehensive evaluation index(ΔE)was used as the objective evaluation standard to verify the test results.RESULTS The training results show that the fitting accuracy of the improved PSO-BP neural network was up to 98.31%.The prediction results show that the prediction error of the improved PSO-BP neural network was the smallest,and the mean absolute percentage error(MAPE),root mean square error(RMSE)and mean color difference(ΔE)were 0.4115,2.1646 and 2.56,respectively.The verification samples of three kinds of particles were prepared for verification.TheΔE of the verification sample and the model drug were 1.73,2.63 and 4.11,respectively.The color difference between the two groups and the model drug was small by visual evaluation.CONCLUSION The BP neural network based on the improved particle swarm optimization algorithm can simulate the prediction of the amount of colorants used in the preparation of Chinese medicine compound granules,and can be used as a recommended optimization model for placebo color matching research.

关 键 词：中药复方颗粒 安慰剂 颜色模拟 神经网络 粒子群算法 CIELab颜色系统 

分 类 号：R283[医药卫生—中药学]