Prediction Model of Salvolatile Column Based on General Regression Neural Networks and Modified Genetic Algorithms

General regression neural networks (GRNN) has a strong ability of approaching non-linear function. It can find the hidden relation between independent variables with dependent variables according to the training sample data. The optimization of the smoothing parameters is crucial to the performance of GRNN, and it is also the essence and difficulty of GRNN training. A modified genetic algorithms (MGA) was applied to optimize smoothing parameters of GRNN, and a model of salvolatile column was built based on GRNN. The model can be applied to predict the carbonization degree and ammonia concentration in the exit of salvolatile column. The proof-testing results indicated that the model possess satisfying predicting performance. Thus, the model of the salvolatile column in this paper can play an important role to stable production.