US EPA

7855413 

Journal Article 

Reaction condition optimization of catalytic hydrogenation of m-dinitrobenzene by BP neural network 

Liu, YX; Wei, ZJ; Chen, JX; Zhang, JY 

2005 

1 

Chinese Journal of Catalysis
ISSN: 1872-2067 

SCIENCE PRESS 

BEIJING 

26 

1 

20-24 

Chinese 

WOS:000227023300006 

Catalytic hydrogenation of m-dinitrobenzene in liquid phase is an attractive routine for the preparation of m-phenylenediamine. To obtain the best result, an artificial neural network was applied to optimize the reaction conditions. A back propagation neural network was trained by the experimental results of homogeneous design using Levenberg-Marquardt algorithm, and the training absolute error was < 1 × 10-5. The trained BP neural network was used to predict the conversion of m-dinitrobenzene and the yield of m-phenylenediamine under different reaction condition schemes. Higher conversion of m-dinitrobenzene was obtained when benzene was used as solvent, but higher yield of m-phenylenediamine was realized when ethanol was utilized as solvent. The most favorable reaction conditions were ethanol as solvent, reaction temperature of 365 K, hydrogen pressure of 2.9 MPa, and catalyst amount of 20%; the yield of m-phenylenediamine was high ≤ 95.8%. The simulation results agreed with experimental data. Thus, the BP neural network could be effectively used to optimize the reaction conditions for the hydrogenation of m-dinitrobenzene. 

neural network; LM algorithm; reaction condition; optimization; m-dinitrobenzene; hydrogenation; m-phenylenediamine