With Si time series of No.7 blast furnace (BF) at Han Iron and Steel Group Co. as sample space, the stationarity of Si time series is tested based on the number of reversed order. A statistical measure of dynamic closeness between two orbits is defined using correlation integral. Divide the Si series into equal-length windows, according to the calculation of the characterization between these windows, the dynamic abrupt changes in BF ironmaking process are detected. Finally the novel DVV (delay vector variance) method is used to examine the predictability of Si time series during the dynamic evolution.