Monitoring and detection with time series models

Many types of random data can be considered as more or less stationary. Stationary stochastic data are characterized optimally by the parameters of a time series model, if model type and model order are known in advance. Recently, a new development in time series analysis gives the possibility to select automatically, with statistical criteria, the model type and the model order for data with unknown characteristics. Hence, the statistically significant features of measured data can be determined without a priori knowledge. This creates the possibility to use estimated and selected models for the automatic monitoring of stochastic data and for the detection of changes. The paper describes variations that can be detected. It shows that considering a measured signal as a stationary stochastic process is already sufficient a priori information to use a powerful statistical framework for the accurate description of observations and for the automatic detection of changes.