双酚A合成反应使用的阳离子树脂催化剂会随温度、时间、负荷和酚酮比的变化而失活;为实现对催化剂活性的在线检测,结合数学回归方法和支持向量机方法,对催化剂失活过程进行混合建模.首先采用数学回归方法描述催化剂活性随使用时间累积而变化的总体趋势,随后在此基础上采用支持向量机拟合温度、负荷和酚酮比对催化剂活性的影响,完成建模过程.运用此方法,针对一个实际的双酚A生产装置建立了生产过程的在线软测量模型,采用该模型对双酚A合成反应过程进行了在线推理估计.仿真及现场运行结果表明,这种基于数学回归和支持向量机混合建模的双酚A合成反应催化剂失活模型的估计精度较高,能够满足实际生产对软测量模型的要求.%During the production of bisphenol A (BPA), the used ion exchange resin catalyst will be deactivated with the changes of reaction temperature, operation time, ratio of phenol to acetone in feed and load of production. In order to monitor the change of catalyst activity online, based on mathematical regression and support vector machine (SVM), a hybrid soft sensing model for catalyst activity of BPA production was established. During the establishment of the above model, a quadratic polynomial was applied for the mathematical regression of the decreasing trend of the catalyst activity; and based on it, the SVM was used for the consideration of the catalyst activity fluctuation caused by the reaction temperature, ratio of phenol to acetone and feed load. Finally a soft sensing model used for the estimation of the BPA content in the reactor outlet of the BPA production was established too. The results of the model simulations and the industrial production process show that the established soft sensing model for the catalyst activity of BPA production can yield satisfactory estimation results and meet the needs of industrial production process.
展开▼
