For objective reasons, machines experience short-term overload, which can lead to damage or destruction of the structure. Under these conditions, it is important to estimate the residual resource of machine fatigue characteristics. For this purpose in engineering practice the method of single overloads (from a high level to low) is widely used. The method allows simulating a mode of adverse irregular loading. Single overload at a level exceeding the initial endurance limit for a certain number of cycles reduces the endurance limit and the residual resource of the structure durability. The empirical dependences proposed in the literature for estimating the relative decrease in endurance limits from the overload coefficient and the cyclic ratio give a complete decrease in the secondary endurance limit at a cyclic ratio equal to one. This is inconsistent with experimental results indicating the existence of a marginal reduction in secondary endurance limits. The formula allowing to correct slightly these dependences and to describe marginal decrease of secondary endurance limits, also under some conditions gives full decrease of secondary endurance limits or loses physical sense. Although the dependence proposed for titanium alloys gives a marginal reduction in the secondary endurance limits other than zero, it determines the anomalous nature of the asymptotic curve which does not correspond to numerous experimental data and mathematical models of damage. Earlier, the author has developed a mathematical model of single overloads and on its basis he has performed an estimation of the residual resource of machine durability based on the results of statistical tests of laboratory samples. In this paper, this model is used to estimate the residual structural strength of machines. The calculated dependences are proposed that satisfactorily describe the experiment and are free from the above shortcomings. These relations can be recommended for implementation in the practice of engineering calculations.