论文部分内容阅读
化工过程数据具有变量多,数据量大的特点,而且在测量过程中易发生数据缺失。为减少数据缺失对数据分析及故障检测过程的影响,需要对缺失数据进行补值。本文采用EM-PCA(ExpectationMaximization algorithmforPrincipalComponentAnalysis)补值算法对TE(Tennessee Eastman)化工过程数据的随机缺失进行补值。选择不同的初值设置方法,并选取不同主元数对不同缺失率下的数据进行补值,应用补值与原始数值的平均相对误差来评价补值结果。结果显示当选用的主元数增大时,补值结果趋于稳定,而且EM-PCA补值算法的误差小于使用平均值法补值及当前值补值方法的误差。补值能够为后续的过程故障检测提供完整的数据,对化工过程的监控具有重要的意义。
Chemical process data with a large number of variables, the characteristics of a large amount of data, but also in the measurement process prone to missing data. To reduce the impact of data loss on data analysis and fault detection processes, missing data needs to be replaced. This paper uses EM-PCA (Supplementary Massaxization algorithmforPrincipalComponentAnalysis) complement algorithm for TE (Tennessee Eastman) chemical process data random missing value. Choose different initial value setting methods, choose different principal components to make up the value of the data under different deletion rates, and apply the average relative error between the initial value and the initial value to evaluate the supplementary result. The results show that when the number of principal components used increases, the value of the revaluation tends to be stable, and the error of the EM-PCA reassignment algorithm is smaller than that of the averaging method and the current value rebuilding method. The replenishment value can provide complete data for the subsequent process fault detection, which is of great significance to the monitoring of the chemical process.