Approach of outlier detection in process control system

In this paper, an outlier detection method using wavelet analysis is proposed in the field of process control system. Such method that is based on a robust model can identify outliers in a time series because of characteristics of wavelet analysis, which is honored as `microscope in math´. In addition, as there is a problem about wavelet analysis, which is the threshold issue, this paper proposes the idea that connect hidden Markov model with wavelet analysis. As hidden Markov model is a statistical model, it can detect outliers directly after analyzing the wavelet coefficients. Experiments are conducted using dataset generated by electric arc furnace mechanism model. The result indicates that the proposed method could detect the outliers effectively.