中文摘要：

分析了熔融拉锥过程的关键工艺参数，指出该过程是一个高度非线性、多变量耦合作用的复杂过程。建立了BP人工神经网络模型并通过程序实现了工艺参数的初选。在此基础上构造的知识库和推理机最终实现了人工神经网络和专家系统的结合。通过该专家系统制作的耦合器实验说明，该系统能够较好地完成工艺参数的优化并实现耦合器制作的自动控制。

英文摘要：

Through the discussion of the key processing parameters in the fused biconical taper process, firstly the paper pointed out that this process was highly nonlinear and multi variable coupled. And then a model of artificial neural network with error back propagation was set up, which was applied to choose the best processing parameters by a computer program. Based on these efforts, both the repository and the inference system were established, which eventually combined the technology of neural network and expert system. The expert system has been proved to be able to optimize the processing parameters and control the process of optical coupler manufacture automatically from its application in experiments.