Abstract: This paper proposes a novel-modeling algorithm based on on-line and off-line parameters fusion and describes the modeling method of two sub models based on adaptive fuzzy neural networks and fuzzy inference. Multiple model fusion-modeling algorithms are designed. For real data of penicillin fermentation process, the simulation results show that modeling accuracy of the algorithm is better.

Key words: Data fusion; Adaptive fuzzy neural networks (ANFIS); Fuzzy inference; Fermentation process; Modeling

摘要： 该文提出了同时利用在线参数和离线参数融合建模的实用新方法，给出了基于自适应模糊神经网络方法和模糊逻辑推理方法的建模过程，将两种建模方法进行最优加权融合，采用真实青霉素发酵过程数据进行模型验证，仿真结果表明了该方法具有较好的建模精度和实用性。

关键词: 数据融合；自适应模糊神经网络；模糊推理；发酵过程；建模