The Neural Network Is Applied Research In Green Chemistry In The Synthesis Reaction

Since 1988, ANN began to be applied in the research of chemistry and chemical industry. The people designed the different neural networks to solve the different application problems, in which there were many literatures. But application of ANN method for the green synthesis of adipic acid had been rarely found.The green chemical synthesis is one of domain research directions in chemical industry. Adipic acid is an important fine chemicals, which was widely used for production of the nylon 66 and other fields. Development of production technology of adipic acid has the importance of society and the economy. This new technology would replace the old production process which give off lot of harmful gases.This green reaction process involves many non-linear and the time-variable characteristic correlation laws, some of which are unable to be treated and described with the traditional mathematics methods.ANN methods can be divided into four types of networks in development of recent years. Based on MATLAB software platform, the author uses the foreword -feedback neural network structure and BP network model, to simulate and predict the effect of single factor and multi-factor on the green synthesis. Some problems in the application of ANN method in the green synthesis of adipic acid were analyzed and discussed through several examples using the networks. The number of implicit nerve unit in BP network influences greatly the performance of network. Many researchers have devoted continuously to this aspect research, but not yet obtained the better theory. It was found that there certain relation of number of implicit nerve unit, the input vector and the number of samples by simulation of many examples. A better qualitative selective range for these parameters was obtained on this ANN system. The effects of learning training algorithm and learning speed selected on the network performance were also discussed.The networks were evaluated from three aspects of the error of sample simulation, prediction of unknown data, the network-mapping tendency, in order to analyze the moderateness of network structures.Four example networks(ANN1～ANN4) were set up in this work. Results of the network simulation and the data predicted shown to accord with the principles of chemistry, and also tallies with the experimental results. The calculated data are consistent with results predicted by the other means (linear interpolation, orthogonal analytic method), and can meet the demand of this green synthesis.In summary, the networks designed is reasonable and has better application performance. The methods developed in the dissertation may provide a new method and a way for the study on the green synthesis of adipic acid.