【摘要】 SiC -CVD过程的本质复杂性制约了在碳 /碳复合材料表面有效地制备高性能抗氧化涂层 .本研究在对CVD工艺过程及机理试验研究的基础上 ,采用人工神经网络技术对SiC -CVD过程进行了辨识与模拟研究 ,建立了SiC -CVD过程的神经网络结构模型 ,并根据CVD过程复杂、工艺性强等特点 ,从样本选取、网络结构设计、学习参数调整等方面对神经网络学习算法进行了改进 .结果表明 :模型对工艺参数影响规律的预测结果与工艺实验结果相一致 ,所开发的神经网络模型 ,不仅可以对各种实验条件下的沉积结果进行准确的预测 ,还能更为全面地反映不同工艺因素对沉积规律的不同影响和判断各工艺因素间的交互作用的存在及作用大小 ,同时采用SiC -CVD模型可对沉积机理的时间效应进行预测和分析 .更多还原

【Abstract】 A neural network (NN) model of the gas_phase and chemical kinetics in a silicon carbide chemical vapor deposition (CVD) process were developed. The model, which was learned from the experiment using 21 parameters within 7 experimental factors and adjusted the network structure automatically, can predict the SiC-CVD results correctly under any given experimental condition in a vertical reactor. It can also be used in the determination of the affections among different parameters and in the analysis of changes happened in CVD process as the deposition continues. A good agreement is founded between the model predicted results and experimental results, the predicted results also indicate that gas_phase mass conservation kinetic process may be very important in SiC-CVD process.更多还原