关键词: 支持向量回归, 中药, 浓度, 软测量

Abstract: Aiming at the problem of the estimating of the concentrated-solution’s concentration during Chinese traditional medicine concentration process，a soft sensor modeling method based on support vector machines is proposed.The concentration-process and technology of Chinese traditional medicine are analyzed，and six process-variables are chose as auxiliary ones.A soft sensor model is constructed by SVR method，and the parameters are optimized and checked making use of the process-data.The concentration during the Chinese-traditional-medicine concentration process，which is predicted by the optimized model，validates the study-capability and generalization-ability of the model through the data.It turns out that the soft sensor model can provide an accurate prediction to the concentration of Chinese traditional medicine，and the model which is constructed by support-vector-machines method has advantages of quick computation and better generalization ability if used to small samples-learning.

Key words: support vector regression, Chinese traditional medicine, concentration, soft-sensor