【摘要】 青霉素发酵过程是一个严重非线性、时变、复杂的动态过程,发酵过程中一些关键参数(如菌体质量浓度ρX、基质质量浓度ρS和产物质量浓度(ρP)难以通过常规仪表在线测量,这些参数的获取非常耗时和困难。提出一种基于粒子群模糊神经网络逆(PSO-FNN逆)的软测量方法。首先给出青霉素发酵过程数学模型,然后根据逆系统理论证明其可逆性,在此基础上构建基于PSO-FNN逆的青霉素发酵软测量模型,最后通过仿真验证该方法性能。仿真结果表明:PSO-FNN逆的青霉素发酵软测量方法能够结合基于发酵机制和纯数据驱动2种软测量方法的优点,对不直接可测的关键参数实现在线软测量,较PSO-BPNN逆和PSO-BPNN软测量方法具有更高的预测精度和更强的预测能力。更多还原

【Abstract】 Penicillin fermentation process was usually a seriously nonlinear,time-varying and complex process,thus it was very difficult to measure the key parameters(such as cell mass concentration ρX,substrate mass concentration ρS and penicillin mass concentration ρP) online using conventional instruments.The obtaining of these parameters was time consuming and difficult.A soft-sensing method based on particle swarm optimization fuzzy neural network(PSO-FNN) inversion in fermentation process was proposed.Firstly,the mathematical model of penicillin fermentation was gained and its reversibility was proved.Then,the PSO-FNN inverse model of penicillin fermentation was constructed.Finally,the method performance was verified by simulation.Simulation results showed that the method could measure the key parameters which couldn′t be measured on line during the course of penicillin fermentation.The method had higher precision and better performance than the method based on PSO-BPNN inversion and PSO-BPNN.更多还原