化工原料中对羧基苯甲醛含量的软测量建模

Soft Sensor Modeling of the 4-CBA Content in Chemical raw Materials

研究化工原料生产中对羧基苯甲醛含量预测问题。传统对4-CBA含量的实时控制的算法依赖样本数据的个数,导致求解复杂,计算速度慢,准确度低。为了克服传统算法的不足,解决PTA氧化过程非线性、时变、大滞后,样本数据采集困难等问题,提出了一种基于最小二乘支持向量机的软测量建模方法,采用小样本的函数回归预测建立了对二甲苯(PX)氧化反应器的数学模型,通过交叉验证的方法进行参数选择,结果表明预测模型具有良好的非线性,较高的精确度,泛化能力强,是用于PTA氧化过程中4-CBA含量预测的一种有效的方法。

Carboxy benzaldehyde content prediction in production of chemical raw materials is researched. In or- der to control the 4-CBA content in real-time, traditional algorithms rely on the number of sample data, resulting in solving complication, slow computing speed and low accuracy. In order to overcome the deficiencies of traditional algorithms and solve the problems of nonlinear, time-varying, time delay, the difficulty of sample data collection in PTA oxidation process, a soft measure based on least square support vector machines （LS-SVM） is proposed. A mathematical model of P-xylene（PX） oxidation reactor is established through mall sample function regression predic- tion. Cross validation method is used to select parameters. The result shows that the prediction model has good non- linear, high accuracy, and generalization ability, so it is an effective method in forecasting of 4-CBA content of PTA oxidation process.