论文部分内容阅读
为了提高乙醇固态发酵过程在线监测的精度,开展了基于傅里叶近红外光谱(FT-NIRS)分析技术的乙醇固态发酵过程参数快速定量检测研究。采用联合区间偏最小二乘法(siPLS)对标准正态变量变换(SNV)预处理后的光谱进行特征波长区间优选;引入遗传算法(GA)、竞争自适应重加权采样(CARS)法和迭代保留信息变量(IRIV)法从优选后波长区间中进一步筛选特征波长变量;最后,建立不同变量筛选方法所得特征波长的乙醇固态发酵过程参数(乙醇和还原糖含量)的偏最小二乘(PLS)预测模型。实验结果显示,与GA和CARS方法相比,IRIV方法所得的波长变量数最少;其中,与乙醇和还原糖相关的特征变量个数分别为43和40;在验证集中,PLS预测模型乙醇含量的验证集均方根误差(RMSEP)和预测相关系数Rp分别为0.2511和0.9934,还原糖含量的RMSEP和Rp分别为0.1730和0.9926,其预测精度亦高于其他方法所得结果。实验结果表明,利用近红外光谱分析技术实现乙醇固态发酵过程关键参数的在线检测是可行的;并且IRIV方法是一种有效近红外光谱特征波长优选方法,可提高预测模型精度。
In order to improve the on-line monitoring accuracy of ethanol solid-state fermentation process, a fast and quantitative detection of ethanol solid-state fermentation process based on FT-NIRS was carried out. The spectral interval preconditioned by Standard Normal Variables Transform (SNV) was optimized by the joint interval partial least squares (siPLS) algorithm. Genetic algorithm (GA), competitive self-weighting weighted sampling (CARS) The information variable (IRIV) method was used to further screen the characteristic wavelength variables from the preferred wavelength range. Finally, partial least squares (PLS) prediction of ethanol solid state fermentation process parameters (ethanol and reducing sugar content) model. The experimental results show that the IRIV method has the least number of wavelength variables compared with the GA and CARS methods. Among them, the number of characteristic variables related to ethanol and reducing sugar are 43 and 40 respectively. In the validation set, the PLS prediction model ethanol content The root mean square error of validation (RMSEP) and predicted correlation coefficient Rp were 0.2511 and 0.9934, respectively. The RMSEP and Rp of reducing sugars were 0.1730 and 0.9926, respectively. The prediction accuracy was also higher than those obtained by other methods. The experimental results show that it is feasible to detect the key parameters of ethanol solid state fermentation by near infrared spectroscopy. And the IRIV method is an effective wavelength selection method of near infrared spectrum, which can improve the accuracy of the prediction model.