论文部分内容阅读
针对发酵过程强烈的非线性和时变性特点,提出一种基于卡尔曼滤波器(KF)和多向核主元分析(MKPCA)的方法对发酵过程进行在线监控。该方法将三维数据空间按批次方向展开为二维数据空间并进行标准化,之后采用KPCA方法获取正常间歇过程的非线性特征,建立更为精确的过程监控模型。在新批次反应过程中利用卡尔曼滤波器对当前批次的未来测量数据进行实时估计从而实现在线监控。该方法和传统MPCA方法的监测性能在一个青霉素发酵仿真系统上进行了比较。仿真结果表明:该方法具有更好的监测性能,能有效获取过程变量之间的非线性关系,降低运行过程的误报率,且能较早检测出过程存在的故障。
Aiming at the strong nonlinear and time-varying characteristics of fermentation process, a method based on Kalman filter (KF) and multi-directional kernel principal component analysis (MKPCA) was proposed to monitor the fermentation process online. The method expands and normalizes the three-dimensional data space into two-dimensional data space in batches, and then uses the KPCA method to obtain the nonlinear characteristics of the normal intermittent process and establishes a more accurate process monitoring model. In the new batch reaction process using Kalman filter on the current batch of real-time measurement of future measurements in order to achieve online monitoring. The monitoring performance of this method and the traditional MPCA method is compared on a penicillin fermentation simulation system. The simulation results show that this method has better performance of monitoring, can effectively obtain the nonlinear relationship between process variables, reduce the false alarm rate during operation, and can detect the process failure earlier.