Development of Multivariate Control Chart Support Vector Data Description Based on Adaptive Exponentially Weighted Moving Average (AEWMA)

Statistical process control (SPC) plays an important role in monitoring and improving product quality. To continuously monitor the product quality, one method that is commonly employed in SPC is the control chart. However, in the real monitoring data process, it is difficult to find the pattern of monitored quality characteristics. To overcome this issue, the support vector data description (SVDD) based control chart can be used when the distributions of quality characteristics are relatively varied or even unknown. Along with the simplicity of implementing SVDD on the control chart, this method has drawbacks when a fluctuating process shift occurs. The adaptive exponentially weighted moving average (AEWMA) control chart can be applied for such a case. Therefore, in this paper, the integration between SVDD and AEWMA is proposed to overcome the limitations of conventional multivariate charts. The performance of the proposed chart is evaluated and compared with the combination of SVDD and exponentially weighted moving average (EWMA) chart using the synthetic dataset. Based on the simulation studies, the proposed method is able to minimize the possibility of false alarms and to provide better performance when there is a large or even small process shift. In addition, an illustration of the application of this method is provided with the real example case.