This model makes use of the interface between the fault detection and fault isolation, which is theoretically employed by fault characteristic matrix. The matrix records the information of correlation between the detection signal and the fault as well as corresponding time. The state characteristics of the system can be reasoned from the relation between the fault signal and the fault. We can make full used of fault signals. The occurring time of the fault can be got by conducting the discrete time method on the observed fault signal. The application of the proposed detection method to the network of mine water level sensor proved that this method is more reasonable than the standard binary fault detection method.