基于加权RBF神经网络的诺西肽发酵过程菌体浓度软测量

Weighted RBF neural network based soft sensor of biomass in Nosiheptide fermentation process

结合诺西肽发酵过程的实际情况,提出了基于加权RBF神经网络(weighted RBF neural network,WRBFNN)的菌体浓度软测量建模方法。在诺西肽发酵过程非结构模型的基础上,根据隐函数存在定理确定出辅助变量,从而使其选择有严格的理论依据。针对菌体浓度变化范围大这一特点,将传统RBF神经网络(RBFneural network,RBFNN)的误差函数进行了改进;然后根据每批训练样本对被预测对象的预估能力,自适应地为各个批次的训练样本分配权重,进而实施WRBFNN建模。实验结果验证了所提方法的有效性。

Combined basis function neural on the unstructure according to the d with the practical situation of Nosiheptide fermentation process, a weighted radial network （WRBFNN） based biomass soft sensor modeling method is presented. Based model of Nosiheptide fermentation process, the secondary variables were selected mplicit function existence theorem, which made the selection theoretically strict. According to the characteristics that biomass could vary in a wide range, the error function of the traditional RBFNN was improved. Each batch sample was self-adaptively weighted according to their predicting ability to the predicted object, and then WRBFNN was used to develop the biomass soft sensor model. The testing results showed the effectiveness of the presented approach.