动物细胞悬浮培养过程神经网络逆解耦控制

动物细胞的悬浮培养以细胞增殖快、生产效率高等优势,成为动物细胞大规模培养的首选方式。而动物细胞悬浮培养过程是一个非线性、强耦合的多输入多输出系统,对一些生物参数(如细胞密度、基质浓度和产物浓度)的控制是提高整个生产水平的关键,应用神经网络逆系统方法对动物细胞悬浮培养过程进行线性化解耦控制,根据培养过程的特点,给出了相应的数学模型,并证明了系统的可逆性,利用神经网络的非线性逼近能力辨识出原系统的逆系统,然后串接在原系统前面构成伪线性复合系统,使动物细胞悬浮培养过程线性化解耦成三个子系统：一阶线性细胞密度子系统、一阶线性基质浓度子系统和一阶线性产物浓度子系统,最后设计模糊PID控制器对各解耦后的线性子系统进行控制,避免了传统PID控制器最优参数选取困难的问题。仿真结果表明,神经网络逆系统方法实现了对动物细胞悬浮培养过程的线性化解耦,系统对给定输入实现了高性能跟踪控制。

Neural network inverse decoupling control of animal cell suspension culture process

The suspension culture of animal cells is the preferred choice for large-scale culture of animal cells because of its advantages of fast cell proliferation and high production efficiency. The process of animal cells suspension culture is a multi-input and multi-output system with its nonlinear and strong coupling. The control of some biological parameters, such as cell density, matrix concentration and product concentration, is the key to improving the whole production level. The method of neural network inverse system was applied to control the suspension culture of animal cells. According to the characteristics of the culture process, the corresponding mathematical model is given and the reversibility of the system is proved. The inverse system of the original system is identified by the nonlinear approximation of neural network, then it is connected in front of the original system to form a pseudo linear composite system. Thus the process of animal cell suspension culture is linearly decoupled into three subsystems: first order linear cell density subsystem, first order linear matrix concentration subsystem and first order linear product concentration subsystem. Finally, the fuzzy PID controller is designed to control the decoupled linear subsystems, which avoids the difficulty of selecting the optimal parameters of the conventional PID controller. The simulation results show that the method of neural network inverse system can realize the linearization and decoupling of the process of animal cell suspension culture, and the system can realize high performance tracking control for the given input.