基于BP网络模型的水合肼合成收率预测及参数优化

Analysis of Yield and Optimization of Parameters Based on BP Neural Network for Hydrazine hydrate Synthesis

采用BP算法建立了水合肼合成收率的神经网络模型,所建网络有良好的回想和泛化性能,对水合肼合成收率预测的平均相对误差小于0.5%。运用网络模型考察了反应温度、物料配比、次氯酸钠流量等操作参数对水合肼合成收率的影响,同时确定了获得较高合成收率时各操作参数的优化值。

A back-propagation （BP） neural network model is established to predict the yield and optimise the operation parameters for Hydrazine hydrate synthesis. The results show that the model based on operation data can predict the yield accurately with the average relative error is less than 0. 5%. The effects of reaction temperature and mixture ratios on the yield are discussed and the optimum values of operation parameters for high yield of Hydrazine hydrate synthesis are obtained.