基于多动态核PCA的统计过程监测策略研究

摘要

针对工业生产过程数据多具有动态和非线性等特性，在时序相关性和非线性理论的基础上提出了一种多动态核PCA的监测方法。该方法突破了MKPCA单模型、非动态的建模方式，构造了适合批量生产过程的多模型、非线性和动态的建模方法，并侧重于在线过程性能监测的实时性，消除了预报未来测量值带来误差，提高了过程性能监视的准确性，并通过实例数学模型验证了该方法的有效性。%　　For the dynamics and nonlinearities of industrial production process data, a multi-dynamic kernel PCA monitoring method based on timing correlation and nonlinear theory was proposed. The method broke through single model, non-dynamic modeling approach for MKPCA. The multi-model, nonlinear and dynamic modeling method suitable for batch production process is constructed, which focuses on real-time performance in online monitoring process, eliminates errors brought by forecasting the future measured values and improves the accuracy of the process monitoring performance. The validity of the method is verified by the sample of the mathematical model.