Artificial Neural Network Simulation Of Energetic Materials Prepared By Spray Drying

The desensitization of energetic materials is a hot topic in this field,and spray drying is an effective desensitization method.Researches show that micron sized energetic materials could be continuous obtained by spray drying.However,due to the danger of the experiment and the complexity of the process.Energetic materials(CL-20,RDX and HMX)prepared by spray drying could be simulated by using artificial neural network(ANN),which had greatly reduced the operational risk,and was of great significance to spray drying.The main research contents of this article were as follows:(1)An artificial neural network model for energetic materials prepared by spray drying was established with gas flow,liquid flow,inlet temperature,mass fraction,relative molecular weight and viscosity as inputs and average particle size and particle size distribution as outputs.(2)Artificial neural networks were optimized by different types(Feed-forward back propagation neural network(FFBPNN),Cascade-forward back propagation neural network(CFBPNN),Elman-forward back propagation neural network(EFBPNN),Layers Recurrent neural network(LR)and Nonlinear autoregressive neural network(NARX)),and different algorithms(Levenberg-Marquardt algorithm(L-M),Momentum gradient descent and adaptive learning rate algorithm(GDX),Genetic algorithm(GA)and Particle swarm optimization algorithm(PSO)),to predict the average particle size.The best model was GA-LR.(3)The operators(selection operator,crossover operator and mutation operator)in the GA were optimized to obtain the best parameters: the selection operator was pairwise competition method,the crossover operator was uniform crossover and the mutation operator was uniform mutation,the crossover probability was 0.8 and the mutation probability was 0.1.Based on the Matlab GUI,a prediction interface for the average particle size and particle size distribution is established.(4)By comparing the performance of GA-LR and empirical correlations,the effects of operating conditions(gas flow,liquid flow,inlet temperature,mass fraction and viscosity)on average particle size and particle size distribution were investigated,which would provide a theoretical basis for quantitative calculations.