Abstract: Smoother is one of the widely used approaches to eliminate or to weaken the stochastic error from sampling data. A new kind of fault-tolerant smoothing algorithm is presented in this paper. This so-called fault-tolerant twice-median smoother is constructed with doubly sliding median smoothers and sliding mean smoothers, and can be used to smooth noise and to extract signals without removing any outliers before. Some theoretical results and simulation show that this new smoothing algorithm has the abilities to avoid bad influence from outliers and patchy in sampling time series and to extract signals unfailingly.