This paper presents an experimental investigation of a recently developed Kronecker Product (KP) method to determine the type, location, and intensity of structural change from an identified state-space model of the system. Although this inverse problem appears to be highly nonlinear, the system mass, stiffness, and damping matrices are identified through a series of transformations, and with the aid of the Kronecker product, only linear operations are involved in the process. Since a state-space model can be identified directly from input-output data, an initial finite element model and/or model updating is not required. The test structure is a two-degree-of-freedom torsional system in which mass and stiffness are arbitrarily adjustable to simulate various conditions of structural change or damage. This simple apparatus illustrates the potential applicability of the system identification technique for damage detection problems by not only identifying the location and the extent of the damage, but also differentiating the nature of the damage. The results are successfully confirmed by laboratory tests.