论文部分内容阅读
针对化工过程数据的特点,提出一种基于形态-EMD滤波的过程数据预处理方法,先构造一类广义形态滤波器滤除过程数据的粗差干扰,引入经验模式分解滤波法消除随机噪声成分,降低正常过程数据的波动,提取过程数据特征成分,显著提高过程监控性能。与传统的过程数据滤波方法相比,形态-EMD滤波方法无需预先确定滤波器参数,是一种完全的数据驱动型方法,具有较好的自适应能力。仿真实验的结果表明,对过程数据的滤波预处理可以提高数据信噪比,显著提高故障检测的准确性。
Aiming at the characteristics of chemical process data, a method of process data preprocessing based on morphological -EMD filtering is proposed. First, a kind of generalized morphological filter is constructed to filter out the gross interference of the process data. Empirical mode decomposition filter is introduced to eliminate random noise components, Reduce the fluctuation of the normal process data, extract the characteristic components of the process data and significantly improve the process monitoring performance. Compared with the traditional process data filtering method, the morphological -EMD filtering method does not need to determine the filter parameters in advance. It is a complete data-driven method with good self-adaptability. Simulation results show that the filtering of process data can improve the signal-to-noise ratio and improve the accuracy of fault detection significantly.