Analysis On Robust MEWMA Control Chart Based On FAST-MCD Algorithm

In the research of multivariate statistics process control, the Multivariate exponentially weighted moving average (MEWMA) control chart with the time domain features is more efficient than the T~2 control chart for detecting small shifts in the multivariate quality process. The sample mean vector and covariance matrix is always used as the estimator of the population mean vector and covariance matrix in the computation of MEWMA statistic. When the outlier data exists in the initial sample data, the signal probability may be distorted for the capture effect.The FAST-MCD algorithm is discussed and D-D chart is used to analyze the influence of algorithm results by the alteration of subsample h. The optimization method of the extraction proportion ? can raise the accuracy of FAST-MCD algorithm compared with the fixed h value and get more robust sample mean vector and covariance.To reduce the impact of capture effect, the FAST-MCD algorithm is used to build a robust covariance matrix to improve the robustness of MEWMA control chart. From the analysis of samples, the robust MEMWA control chart based on FAST-MCD algorithm has better immunity to small amount of noise in the initial samples.