Abstract: Traditional fuzzy neural network lacks simplicity of design and real-time controlling.A self-constructuring double fuzzy neural network algorithm is proposed.In general，the mathematical description of the existing rules can be expressed as a set of clusters and the fuzzy membership functions and fuzzy rules are tuned automatically off line.The self-constructuring double fuzzy neural network consists of two fuzzy neural networks.One network is learning and another is controlling during on-line work time;parameters of the double fuzzy neural network are synchronized after one system cycle;the output of the controlling fuzzy neural network is taken as the algorithm’s output.Simulation experiments are carried out on the test bed of rocket engine.The results show that this algorithm has improvement in real-time control and robustness against out-of-range parameters，and decreases the design complexity of the fuzzy neural network.

Key words: double neural network, self-constructuring fuzzy neural network, rocket engine test bed

摘要： 针对传统模糊神经网络设计复杂、控制实时性滞后的问题，提出自组织双模糊神经网络算法。将样本数据进行聚类划分，形成原始的模糊隶属函数集;在神经网络的离线训练过程中，完善并优化模糊隶属函数和规则;采用双神经网络结构，在线工作时，一个神经网络完成在线学习任务，另一个神经网络完成工业控制任务;经过一定的系统周期，同步系统中两组神经网络的参数;提取完成控制任务的神经网络的输出作为算法的输出。应用于火箭发动机试验台控制系统中，表明算法能够提升控制系统中针对输入参数越界的鲁棒性，提高控制实时性，简化了模糊神经网络的设计复杂度。

关键词: 双神经网络, 自组织模糊神经网络, 火箭发动机试验台控制系统