A scheme for balanced monitoring and accurate diagnosis of bivariate process mean shifts

Monitoring and diagnosis of mean shifts in manufacturing processes become more challenging when involving two or more correlated variables. Unfortunately, most of the existing multivariate statistical process control schemes are only effective in rapid detection but suffer high false alarm. This is referred to as imbalanced performance monitoring. The problem becomes more complicated when dealing with small mean shift particularly in identifying the causable variables. In this research, a scheme to enable balanced monitoring and accurate diagnosis was investigated in order to improve such limitations. Design considerations involved extensive simulation experiments to select input representation based on raw data and statistical features, recognizer design structure based on individual and synergistic models, and monitoring-diagnosis approach based on single stage and two stages techniques. The study focuses on correlated process mean shifts for cross correlation function, ρ = 0.1 ~ 0.9 and mean shift, μ = ± 0.75 ~ 3.00 standard deviations. Among the investigated designs, an Integrated Multivariate Exponentially Weighted Moving Average with Artificial Neural Network scheme gave superior performance, namely, average run lengths, ARL1 = 3.18 ~ 16.75 (for out-of-control process) and ARL0 = 452.13 (for in�control process), and recognition accuracy, RA = 89.5 ~ 98.5%. The proposed scheme was validated using an industrial case study from machining process of audio-video device component. This research has provided a new perspective in realizing balanced monitoring and accurate diagnosis of correlated process mean shift