In fact, there widely exists a kind of time series data that is non-stationary but can be transformed into several local stationary time series data, the identification problem of the non-stationary time series data is studied, and then this paper proposes a piecewise recursive identification algorithm with mechanism. Based on the definition of stationary time series data, the precipitation algorithm which has the mean variance point, the mutation point and the mutation point of the autocorrelation function, is constructed based on the statistical characteristics such as the mean and variance and autocorrelation function. On this basis, a series of locally stationary sub sequences are identified from the identified non-stationary sequences, and then, the Burg algorithm is applied to the recursive identification of local stationary subsequences. The experimental results show that the new algorithm can divide the boundary points of the local stationary sub sequences with smaller position deviation. At the same time, the calculation efficiency is improved significantly under the condition of high accuracy.