基于非线性核脊回归算法的海洋微生物发酵参数软测量

针对海洋微生物溶菌酶发酵工艺中参数实时测量,采用根据发酵过程中主辅变量关系的相关灵敏度矩阵绝对值最大元素排序选择辅助变量,运用非线性核脊回归(NKRR)算法预测主导参数的软测量方法.相关灵敏度矩阵反映主辅变量联系,根据发酵实际实验得到.改进直接优化的非线性核脊回归算法辨识发酵非线性系统,将空间Rn中主导变量和辅助变量非...

Soft-sensing of Microorganism Ferment Parameters Based on Non-linear Kernel Ridge Regression Algorithm

Regarding the real-time measurement of parameters in lysozyme zymotechnics of marine microorganism,a soft-sensing method was adopted.This method selected the auxiliary variable upon the correlation sensitivity matrix which reflecting the relation between the main and auxiliary variable in fermentation,and employed the non-linear kernel ridge regression（NKRR） algorithm to predict the main parameters;and thus the nonlinear function mapping relation between the main and auxiliary variable in space Rn was converted to linear division in space H,and by using Kernel Method of Mercer to solve the regression equation,the direct calculating of the variable mapping function and its inner product were averted.By simulating and comparing SVM and RBFNN algorithms,the results prove the excellent generalization performance of this improved soft-sensing method,and both real-time and accuracy can meet requirements.