摘要
针对生化反应过程中软测量模型随着时间的推移而出现的模型老化现象,提出一种基于增量学习的自适应模糊支持向量机软测量建模方法。它首先将输入空间中的样本映射到高维特征空间,然后根据样本偏离超平面的程度赋予不同的模糊隶属度,建立模糊支持向量机软测量模型,并在模型投入现场运行后,通过一种改进的增量学习算法在线更新模型参数以自适应获得更加准确的软测量模型。以L-赖氨酸流加发酵过程为例,验证了所提算法能够从过程的第2批次开始对关键生物量参数(菌丝浓度和基质浓度)进行较准确的在线预测,与普通的模糊支持向量机建模方法相比具有较高的预测精度和自适应性。
In order to overcome the emergence of model aging as time moves on,a soft sensing modeling method based on incremental learning and fuzzy support vector machines is presented.Data samples in input space are mapped into high dimensional feature space.The fuzzy membership value for each input point is computed according to its distance to the hyperplane,and a soft sensing model based on fuzzy support vector machines is established.In addition,after the model is put into application,the model can be updated on-line through an improved incremental learning algorithm.Simulation experiment on a fed-batch L-lysine fermentation process shows that the crucial biological parameters can be predicted starting from the second batch.Experiment results also show that the proposed method is more accurate and adaptive compared with alternative modeling methods.
出处
《仪器仪表学报》
EI
CAS
CSCD
北大核心
2011年第2期469-474,共6页
Chinese Journal of Scientific Instrument
基金
国家"863"计划(2007AA04Z179)
国家"863"项目子项(2007AA091602)
高等学校博士学科点专项基金(20070299010)资助项目
关键词
自适应学习
模糊支持向量机
软测量
L-赖氨酸发酵过程
adaptive learning
fuzzy support vector machine
soft sensing
L-lysine fermentation process