In combination with unascertained mathematical theory, a unascertained filtering(UF) method is brought forward and the related parameters are investigated. Consequently, the latent problems are analyzed and some corresponding measures advanced. The value of λis optimized by alternative  calculation and the sectional parameter  m determined by enumeration, so the    validity and reliability of this method may be ensured. By using the UF method in the monitoring and experiment,  the problems on error hiding, normal data discarding and hard to differentiate errors from abnormality existing in traditional    methods for detecting errors are settled. The UF method possesses the characters    of dynamic and real -  time, and has the predominance in solving data error series with a long observation period and a wide span obtained from an ERD.  This new approach to process error is feasible and rather ideal.