基于支持向量机的聚酯粘度在线软测量

Data Modeling Based on Support Vector Machine in Soft Sensor

为了找出PET产品的聚酯粘度检测的操作条件与质量指标之间的关系,采用支持向量机对聚酯粘度进行软测量建模。支持向量机是一种基于统计学习理论的新的机器学习方法。进一步研究了容许误差、参数和核函数对建模结果的影响,提出一种模型增量校正方法。仿真结果表明,该方法能够大大提高模型精度。

Conglutination is one of the most important index of PET quality, however detecting instrument can not reflect the fluctuation of quality promptly. For finding the relationship of operation conditions and index of quality, support vector machine is used for modeling the PET conglutination. Support vector machine is a kind of new machine learning method based on statistic. The affection of the permission error parameter and kernel function is considered. A method of adjusting the model on-line is presented, which verifies the model according to the changing data, and simulation result shows that it greatly improves precision of the model.