【摘要】 为提高对发酵过程中质量变量的预测精度,解决发酵数据非线性的问题,提出一种基于核二次互信息回归的质量预测模型。将非线性过程数据核映射至高维特征空间,使其线性可分;基于高维特征空间,使用Renyi二次熵与二次互信息定义目标函数提取过程特征,建立过程特征与质量变量间的回归模型;二次互信息可衡量变量间的非线性关系。仿真实验及大肠杆菌发酵生产数据的实验结果表明,该方法具有较高质量预测精度,对非线性数据有较强处理能力。更多还原

【Abstract】 To improve the prediction accuracy of quality variables for fermentation processes and solve the non-linear problem of fermentation data,aquality prediction model based on kernel quadratic mutual information regression was presented.The nonlinear process data were transformed into a linearly separable form by using kernel projection,resulting in high-dimensional features.Supervised feature extraction for process data was carried out based on high-dimensional space,the objective function defined by Renyi’s quadratic entropy and quadratic mutual information was utilized.A regression model between the process features and quality variables was established.Quadratic mutual information was able to measure the nonlinear dependency between variables.Experimental results of the simulation platform and the production data of E.coli fermentation show that the proposed method has higher prediction accuracy and better nonlinear-processing ability.更多还原