论文部分内容阅读
采用径向基神经网络(RBFNN)结合近红外光谱(NIRS)技术建立一种分析安络小皮伞发酵菌丝体中甘露醇、多糖和腺苷三种组分的定量分析模型。收集164个安络小皮伞液体发酵菌丝体样本的近红外光谱数据,采用常规方法分别测定样本中甘露醇、多糖和腺苷的含量。在应用蒙特卡罗偏最小二乘法(MCPLS)识别异常样本、确定校正集样本数量的基础上,以逼近度(Da)为评价指标,采用可移动窗口径向基神经网络(MWRBFNN)筛选特征波长变量,筛选最佳光谱预处理方法、隐含层节点数(NH)等模型参数。建立甘露醇、多糖和腺苷组分定量分析模型,最佳RBFNN-NIRS模型中校正集和预测集样本实验测定值与预测值间相关系数分别为0.9274、0.9009、0.9440和0.9354、0.9018、0.8847,表明模型具有很好的拟合度和预测性能。
A RBFNN combined with near infrared spectroscopy (NIRS) technique was used to establish a quantitative model for the analysis of three components of mannitol, polysaccharides and adenosine in the mycelium of Rhizoctonia solani. Nearly infra-red spectrum data of liquid ferment mycelia of 164 paraquat colonies were collected, and the contents of mannitol, polysaccharides and adenosine were determined by routine methods. Based on the recognition of abnormal samples by MCPLS and the determination of the number of samples in the calibration set, the approximation degree (Da) was used as evaluation index, and the characteristic wavelength was screened by using MWRBFNN (Moving Radial Basis Function Neural Network) Variables, screening optimal spectral pretreatment methods, hidden layer nodes (NH) and other model parameters. The quantitative analysis models of mannitol, polysaccharide and adenosine were established. The correlation coefficients between the experimental set value and the predicted value in the optimal RBFNN-NIRS model were 0.9274,0.9009,0.9440 and 0.9354,0.9018,0.8847 respectively, It shows that the model has a good fit and predictive performance.