动物细胞悬浮培养过程动态RVM软测量及实现

发布时间：2018-03-18 06:21

本文选题：悬浮培养　切入点：软测量　出处：《江苏大学》2017年硕士论文　论文类型：学位论文

【摘要】：畜牧养殖业迅猛发展,禽流感、口蹄疫等动物传染病随之暴发流行,兽用疫苗的市场需求量大幅度增加,采用传统转瓶培养方式生产疫苗已远远不能满足现代畜牧养殖业的发展。动物细胞悬浮培养技术逐步代替传统转瓶培养方式,并成为当前工业生产生物疫苗的主流技术,所以,如何应用动物细胞悬浮培养技术实现疫苗的快速高效生产成为当前以及未来生物制药领域的焦点和竞争核心。动物细胞悬浮培养过程具有高度的非线性、时变性等特性,过程中由于受到传统物理仪表测量方法的限制(易染菌、成本高等),一些关键状态变量(如葡萄糖浓度、乳酸浓度、细胞密度等)的实时监测严重制约了产物合成从而影响了生物制品的产量和质量。而离线测量又具有一定的滞后性,严重影响了整个过程的优化控制。软测量技术的出现为解决上述问题提供了一种有效途径。本文针对BHK-21(乳仓鼠肾细胞)悬浮培养过程关键状态变量难以实时测量的问题,提出一种基于直角三角形角度值动态加权的软测量建模方法。首先结合培养过程工艺机理,确定模型的主导变量,然后基于现场采集数据,应用相关系数法确定模型的辅助变量,最后建立软测量模型。在深入了解支持向量机、最小二乘支持向量机和关联向量机的基础上,将其与基于直角三角形角度值动态加权的关联向量机建模方法进行对比。结果表明,相比支持向量机与最小二乘支持向量机,关联向量机的解具有较好的稀疏性和更高的鲁棒性,且模型预测精度较高;动态关联向量机不仅具有关联向量机的优点,而且能将工业过程的动态特性引进模型中,使得预测效果更加精确、更能体现工业过程的本质。为了实现BHK-21细胞悬浮培养过程实时监测及优化控制,充分利用MATLAB强大的计算能力和WinCC强大的脚本编程能力,建立基于MATLAB和WinCC混合编程平台的BHK-21悬浮培养过程关键状态变量的在线预测及实时监控系统,为培养过程的优化控制奠定了基础。实验结果表明,所建监控系统能够实时准确的进行关键状态变量的显示,为进一步的优化控制提供了充分的工况信息,降低了系统故障概率,从而实现经济高效的工业生产。
[Abstract]:With the rapid development of animal husbandry, avian influenza, foot-and-mouth disease and other animal infectious diseases have exploded, and the market demand for veterinary vaccines has increased substantially. It is far from satisfied with the development of modern animal husbandry. Animal cell suspension culture technology has gradually replaced the traditional flask culture method, and has become the mainstream technology in the current industrial production of biological vaccine. How to use the technology of animal cell suspension culture to realize the rapid and efficient production of vaccine has become the focus and competition core in the field of biopharmaceuticals at present and in the future. The process of animal cell suspension culture has the characteristics of high nonlinearity, time-varying and so on. Due to the limitation of traditional physical instrument measurement methods (easy to infect bacteria, high cost etc.), some key state variables (such as glucose concentration, lactic acid concentration, etc.). The real-time monitoring of cell density has seriously restricted the synthesis of the product and affected the production and quality of biological products. The emergence of soft sensing technology provides an effective way to solve the above problems. In this paper, the key state variables of BHK-21 (neonatal hamster kidney cell) suspension culture process are difficult to measure in real time. A soft sensor modeling method based on dynamic weighting of angle value of right triangle is proposed. Firstly, the dominant variables of the model are determined by combining the process mechanism of culture process, and then the data are collected on the spot. The correlation coefficient method is used to determine the auxiliary variables of the model. Finally, a soft sensor model is established. On the basis of in-depth understanding of support vector machine, least square support vector machine and association vector machine, It is compared with the modeling method of correlation vector machine based on dynamic weighting of angle value of right triangle. The results show that compared with least squares support vector machine, support vector machine is better than support vector machine. The solution of association vector machine has better sparsity and higher robustness, and the prediction accuracy of the model is higher. Dynamic association vector machine not only has the advantage of association vector machine, but also can introduce the dynamic characteristics of industrial process into the model. In order to realize the real-time monitoring and optimization control of BHK-21 cell suspension culture process, we can make full use of the powerful computing power of MATLAB and the powerful scripting ability of WinCC. The on-line prediction and real-time monitoring system of the key state variables of BHK-21 suspension culture process based on MATLAB and WinCC hybrid programming platform is established, which lays a foundation for the optimization control of the culture process. The experimental results show that, The system can display the key state variables in real time and accurately, provide sufficient information for further optimization control, reduce the probability of system failure, and realize economic and efficient industrial production.
【学位授予单位】：江苏大学
【学位级别】：硕士
【学位授予年份】：2017
【分类号】：Q813.11;S859.797;TP277

【参考文献】