【作者】 陈曦；

【导师】 刘涛； 尹毅强；

【作者基本信息】 大连理工大学 ， 电子信息， 2024， 硕士

【摘要】 1,3-丙二醇(1,3-PDO)作为一种重要的化工原料,在聚合物合成和高分子材料制备中发挥着关键作用。其生物发酵法生产因环保特性而备受瞩目。然而,对该过程分析主要基于采集样本做离线检测,这种方式存在诸多缺点且无法进行在线监测。近红外(NIR)光谱检测技术是一种无损检测技术,具有快速高效、安全无损、操作简单且易于实现在线监测等优点,在生物发酵工程监测中展现出巨大潜力。尽管近红外光谱在生物发酵过程监测中的应用日益增多,但针对1,3-PDO发酵过程的监测研究仍显不足,有待探索。本文以1,3-PDO发酵过程为研究对象,开展了相关研究工作,主要研究内容和贡献如下:首先,以丁酸梭菌发酵生产1,3-PDO体系为对象,搭建了基于近红外光谱检测技术的在线监测实验平台。为实现对底物甘油、产物1,3-PDO、副产物丁酸含量及生物量(OD)进行在线监测,提出了一种基于概率局部重建(PLR)策略的半监督支持向量回归机(SSSVR)光谱标定建模方法,利用协同训练框架进行半监督学习,利用PLR模型估计无标签数据及其伪标签的不确定性,并将其作为置信度选择标准。通过实验验证,所建立的在线监测模型对底物甘油浓度、产物1,3-PDO浓度、副产物丁酸浓度和生物量的预测均方根误差分别为1.2712g/L、0.9872g/L、0.1475g/L和0.2424,满足在线监测的精度要求,可以用于实际环境中监测1,3-PDO发酵过程关键组分含量。其次,针对实时监测1,3-PDO发酵过程关键组分含量的问题,提出了一种基于半监督学习策略的近红外光谱标定建模方法。为解决近红外光谱数据维度高、信息重叠严重的问题,提出了一种二元蜉蝣优化算法(BMA)来选择光谱的特征波长。针对有标签数据不足的问题,提出了一种基于期望最大化(EM)算法的鲁棒半监督概率主成分回归(RSSPPCR)建模方法,充分利用全部有标签和无标签的光谱数据,并且将学生t分布引入模型以提高对离群数据点的容忍度,从而保证同时检测发酵过程多组分含量的可靠性。通过克雷伯氏杆菌厌氧发酵甘油产生1,3-PDO实验数据集,表明该方法能够实时准确地预测1,3-PDO发酵过程中底物甘油、产物1,3-PDO的含量和生物量。最后,为了便捷1,3-PDO发酵过程数据处理与预测模型构建,设计了一款可视化界面,该界面集成了登录、光谱数据处理、校正模型构建以及离线预测等功能模块,该界面的设计极大地简化了数据处理与建模过程,提高了工作效率,具备良好的可扩展性。更多还原

【Abstract】 1,3-propanediol(1,3-PDO),as an important chemical product,plays a crucial role in polymer synthesis and the preparation of high-performance polymeric materials.Its production through biological fermentation has attracted much attention due to its environmental friendliness.However,the analysis methods of this process are mainly based on taking samples for off-line measurement,which has many drawbacks and cannot be used for on-line monitoring.The near-infrared(NIR)spectroscopy takes advantage of fast and non-destructive operation for on-line monitoring and thus is increasingly used for on-line analysis and detection in biological fermentation engineering.Despite the increasing application of NIR spectroscopy in biological fermentation process monitoring,research on monitoring the fermentation process of 1,3-PDO remains insufficient and warrants further exploration.Focusing on the fermentation process of1,3-PDO,this paper conducted relevant research work.The main research contents and contributions are outlined as follows:Firstly,focusing on the 1,3-PDO fermentation system using Clostridium butyricum as the fermenting microbial community,an on-line monitoring experimental platform based on NIR spectroscopy technology is constructed.To achieve on-line monitoring the content of glycerol(substrate),1,3-PDO(product),butyric acid(by-product),and biomass(OD),a semi-supervised support vector regression machine(SS-SVR)model based on probabilistic local reconstruction(PLR)strategy is proposed.The co-training framework is used for semi-supervised learning.The PLR model is utilized to estimate the pseudo label of unlabeled data and the uncertainty of the pseudo label.The uncertainty of the pseudo label is used as the confidence selection criterion.The experimental results show that the established on-line monitoring model had root mean square errors of 1.2712 g/L for glycerol concentration,0.9872 g/L for 1,3-PDO concentration,0.1475 g/L for butyric acid concentration,and 0.2424 for biomass,meeting the requirements of on-line monitoring accuracy.Thus,it can be applied to practical environments to monitor the contents of key components in the 1,3-PDO fermentation process.Secondly,concerning the real-time monitoring of key component contents during a fermentation process of 1,3-PDO,a near-infrared(NIR)spectral calibration modeling method is proposed based on a semi-supervised learning strategy.To solve the problems of high dimensionality and serious information overlap of NIR spectra,a binary mayfly optimization algorithm(BMA)is proposed to select the characteristic wavenumbers of the measured spectra for model calibration.To address the issue of insufficient labeled sample dataset,a robust semisupervised probabilistic principal component regression(RSSPPCR)modeling method is presented based on the expectation maximization(EM)algorithm,which could fully utilize all labeled and unlabeled spectral data.The student t-distribution is introduced into the modeling to improve the model tolerance of outliers,so as to ensure the reliability of simultaneous detection of multiple components during the fermentation process.The experimental results show that the proposed model can effectively predict the important contents of glycerol(substrate),1,3-PDO(product),and biomass during a fermentation process of 1,3-PDO.Finally,a user-friendly interface is designed for the convenience of data processing and predictive model construction in the 1,3-PDO fermentation process.This interface integrates functional modules such as login,spectral data processing,calibration model construction,and off-line prediction.The interface greatly simplifies the data processing and modeling processes,improves work efficiency,and possesses excellent scalability.更多还原