CAJ | 학술논문

依据发酵过程的机理和改进的Elman神经网络动态建模原理,提出了一个新的发酵过程建模分批训练算法.通过发酵过程仿真实验,与传统的BP建模算法比较,改进的Elman神经网络建模算法具有收敛速度快、泛化能力强等特点.此外,利用该算法编制的软件可以内嵌到发酵过程监控系统中,实现发酵过程在线建模与状态参量的在线预估.
의거발효과정적궤리화개진적Elman신경망락동태건모원리,제출료일개신적발효과정건모분비훈련산법.통과발효과정방진실험,여전통적BP건모산법비교,개진적Elman신경망락건모산법구유수렴속도쾌、범화능력강등특점.차외,이용해산법편제적연건가이내감도발효과정감공계통중,실현발효과정재선건모여상태삼량적재선예고.
According to the characteristics of fermentation process and the dynamic modeling theory of modified Elman neural network,a new batch training algorithm for fermentation process modeling is proposed.The training and testing of fermentation process simulation experiments show that modified Elman neural network modeling algorithm has specialties that convergence speed is faster and generalization is better,compared with traditional BP modeling algorithm.Furthermore,the software based on the modeling algorithm can be embedded into fermentation process control system,realizes fermentation process online modeling and state parameter online estimating.