The developments in contemporary power engineering are characterized by the increased unit power of the machines. This has led to an increase in the size of individual components, their stress levels, and to the use of new high-strength materials. At the same time, there often exist metallurgical defects in the billets out of which the machine components are produced. In view of this, it becomes increasingly important to analyze in detail and make in-depth calculations for estimating the reliability of machine elements in which metallurgical and operating defects are present. In view of the variety of geometrical peculiarities of the possible defects and the complex forms of machine components, difficulties are encountered in estimating their operational reliability by calculations with the use of the theory of the mechanics of failure [i] and experimental tests of full-scale components [2]. The most preferable, accurate, and operative method is a combined experimental and calculation approach based on the use of the photoelasticity method and the dependences of the mechanics of failure. We illustrate this in two examples in which three methods were applied to determine the failure rpm of steel and epoxy resin disks. These methods are the experimental, the photoelastic, and the one based on the existing (in the given case) analytical solutions. The results of tests of steel 34KhN3M disks of radius R 2 = 0.275 m and thickness t = 0.075 m were given earlier in [3]. The disks had two radially symmetrical cracklike cuts of length a = 0.025 m emerging from the center hole of radius R I = 0.075 m. The average rotational speed of failure of the disks nf = 8350 rpm. The fracture surface of the disks was brittle in nature. Disk models of the following parameters were prepared from optically sensitive materials with a view to check the possibility of determining the failure rpm of steel disks based on data from the photoelasticity method: R= = 0.137 m, R I = 0.0375 m, t = 0.055 m, i.e., the geometrical modeling factor C E = 2. The model cracks were produced in two stages. At first, the models were heated to the "freezing" temperature and a 0.01 m deep cut made with a sharp blade [4]. Thereafter, the models were loaded with centrifugal forces in a centrifuge till the appearance of natural cracks of (1-2)'10 -3 m length. The models were cooled in an annealing mode. Analysis of sections from an annealed model indicated that residual stresses were absent.