The paper discusses the problem of the accuracy of the identification techniques detecting cracks and corroded members in vibrating beam and frame structures. The presence of the fatigue crack usually causes very small changes of the stiffness of the beam elements of the structure. To detect these changes it is necessary to apply the most precisely mathematical detection technique. The identification procedure based on the least squares technique uses finite element models (FEM) of the structure and as the source of information the measured dynamic response and the natural frequencies. The application of the Dynamic Stiffness Matrix (DSM) [I) for the representation of all constraints and modal equations makes it possible to present the identification process in a very accurate and efficient mathematical form. The methoyof d of the detection of structural changes used in the present paper was described in our previous paper (2). The Consistent Mass Matrices (CMM) and Lump Mass Matrices (LMM) are very often used in the identification algorithms. It is shown that application of simplified approaches (CMM and LMM) can result in lower accuracy and poorer convergence of the identification algorithms. However, the application ofCMM mass matrices does not introduce significant errors. The algorithms were tested on simulated numerical data for ten element beam frames.