论文部分内容阅读
针对工业过程数据的多模态、非高斯分布等问题,提出了一种基于自适应邻域参数的局部标准化独立元分析(adaptive local standardized independent componentanalysis,ALSICA)算法,并将该算法应用于过程故障检测中。针对传统的ICA算法未能考虑过程数据的多模态分布问题,引入局部标准化方法建立了局部标准化独立元分析(LSICA)算法。进一步,考虑到局部标准化方法的邻域参数K值选取的问题,提出了一种新的ALSICA算法,该方法基于密度最优的概念选取邻域参数K值,并且K随着数据点变化做出自适应改变。然后建立独立元监控统计变量,进行在线监控。连续搅拌反应釜(continuous stirring tank reactor,CSTR)系统仿真结果验证了ALSICA方法较传统的ICA方法效果更优。
Aiming at the multi-modal and non-Gaussian distribution of industrial process data, an adaptive local standardized independent componentalgorithm (ALSICA) algorithm based on adaptive neighborhood parameters is proposed and applied to process fault checking. Aiming at the problem that the traditional ICA algorithm can not consider the multi-modal distribution of process data, a locally normalized independent element analysis (LSICA) algorithm is introduced. Furthermore, a new ALSICA algorithm is proposed in view of the selection of neighborhood parameter K of local normalization method. The method selects the value of neighborhood parameter K based on the concept of optimal density and makes K as the data point changes Adaptive change. Then establish an independent meta-monitoring statistical variables for online monitoring. The simulation results of the system of continuous stirred tank reactor (CSTR) verify that the ALSICA method is more effective than the traditional ICA method.