CAJ | 학술논문

将可预测元分析（ ForeCA）与偏最小二乘法（ PLS）结合用于故障检测，在选取合适的可预测元的基础上，运用偏最小二乘回归，进一步提高模型对系统的预测能力，克服了偏最小二乘回归方法无法反映系统动态时序特性的缺陷，并构造CUSUM统计量和SPE统计量以检测故障是否发生。最后通过TE模型上的仿真实验结果表明：ForePLS方法能有效检测慢漂移等故障。
장가예측원분석（ ForeCA）여편최소이승법（ PLS）결합용우고장검측，재선취합괄적가예측원적기출상，운용편최소이승회귀，진일보제고모형대계통적예측능력，극복료편최소이승회귀방법무법반영계통동태시서특성적결함，병구조CUSUM통계량화SPE통계량이검측고장시부발생。최후통과TE모형상적방진실험결과표명：ForePLS방법능유효검측만표이등고장。
The combination of ForeCA (Forecastable Component Analysis) and PLS(Partial Least Squares) was used in the failure detection; basing on selecting the appropriate forecastable components, the PLS was applied to improving the model’ s predictive ability so that the PLS’ s incapability in reflecting the system’ s dynamic timing characteristics can be overcomed, and both CUSUM statistic and SPE statistic for the failure detection can be constructed.Simulation on TE ( Tennessee Eastman) model proves the effectiveness of the proposed method in detecting the slow drift fault.