基于非线性主成分-广义回归神经网络法评估甲醇合成催化剂活性

甲醇合成过程中,影响甲醇单程收率的因数较多,反应机理十分复杂,难以建立准确的机理模型。本文提出了用非主成分分析方法对输入变量预处理,运用广义回归神经网络的非线性映射能力,建立了甲醇合成单程收率的预测模型,并用此模型对不同时期的甲醇合成催化剂的活性进行评估。实例表明,此模型可对甲醇合成催化剂的活性进行定量评估,对指导甲醇生产具有重要意义。

Activity analysis of methanol synthesis catalyst based on nonlinear main component analysis and generalized regressive neural network model

For methanol synthesis process, there are many factors affect the once through yield of methanol and the reaction principle is very complicated, so it is difficult to build exact reaction model. A prediction model for the once through yield of methanol was built by pre-treating the input variables with nonlinear main component method, and using nonlinear mapping of generalized regressive neural network, and used to evaluate the activity of methanol synthesis catalyst at different life stages. It can be seen from the actual example that this model can quantitatively evaluate the activity of methanol synthesis catalyst, and is useful for guiding methanol production.