One of the most important indicators of reliability of technical objects is their durability. Probabilistic and statistical methods used for group assessment of service life of marine equipment at the stage of its design are unsuitable for individual assessment of a residual resource during operation due to the great values of coefficients of the resource variation and significant discrepancies between the proposed and actual operation models. Improving the reliability of estimates can be achieved by controlling the actual operation and monitoring of diagnostic parameters which characterize degradation of the actual technical state of equipment. The article considers on specific examples the features of predictive extrapolation of time-series observations over diagnostic parameter and application of forecasting method based on the actual history of operational loadings. It is stated that the extrapolation method, in spite of such advantages as using a minimum number of controlled parameters and versatility, has a number of shortcomings that reduce its effectiveness in the evaluation of residual resource of multi-mode objects. To extend the use of extrapolative forecasting there have been proposed special methods of adjusting time series using the Euler method, linear and nonlinear regression models and the estimation approach of residual resource, which provides a preliminary study of physico-chemical deterioration of technical condition of equipment and subsequent measurement of the equivalent hours of the resource using results of the check operation and the hypotheses of linear summation of damage.