In this paper, adaptive threshold generation for robust fault detection is analyzed. Uncertainties due to parameter variations are considered in the system model using an interval model. An interval observer is used as a suitable passive robust strategy to generate an adaptive threshold by propagating uncertainty to the residuals. The design procedure of this observer is implemented via pole placement using linear matrix inequalities (LMI). The optimal threshold is generated by evaluating the worst-case residual's energy evolution either in the time or frequency domain. When the advanced adaptive threshold is used, the minimum detectable fault is characterized. Finally, an application example based on two-degree freedom helicopter will be used to assess the validity of the proposed approach and to check the validity of the derived results.