Soft-Sensing Modeling of a Fermentation Process Through Support Vector Machines and Genetic Algorithms

A soft-sensing model is developed for on-line estimate of biomass concentration based on support vector machines.And genetic algorithms are introduced in selection of model input and the parameters of support vector machines.The purpose is to find out the input characteristic variables which contribute most to the model's estimation result for reducing the number of dimensions of space to input and scope of the problem to solve,thus decreasing the difficulties in computation and training practice.Meanwhile,the decision function can be obtained better to improve the performance of support vector machines by way of readjusting parameters.The training/verifying data of the model are all based on the actual experimental process,i.e.Nosiheptide fermentation.Result shows that soft-sensing model optimized by genetic algorithms is highly beneficial to the estimate of biomass concentration.