A new approach to the control problem for a given technical state of o
bjects by input-output signals is presented. The problem is stated and
solved in deterministic form. The abstract control problem (ACP) is s
tated on the basis of the external extension of the definition, and su
fficient conditions of fault detection for linear dynamic systems are
formulated for the first time. It was proved that the solution of a sy
stem of ACP operator equations is reduced to three variants of similar
ity transformation type, independent of the representation form. The f
irst variant corresponds to the fault detection of a Linear dynamic sy
stem with an invertible similarity transformation (the duplication sch
eme). The second variant corresponds to the similarity transformation
regular by Neumann (the reduced order control scheme). The third varia
nt is the degenerate transformation, which does not provide conditions
of guaranteed control. It is shown that the known fault detection met
hods, i.e., the Luenberger observers, surplus variable, algebraic inva
riants, and compatibility relations methods, correspond to the degener
ate similarity transformation. A complete example of different variant
s of the guaranteed control problem is given.