This article presents a novel approach to predicting the service life of mechanical systems. Forecasting is based on the assumption that the product's behavior under operating conditions should be considered as the system's behavior operating in a blow-up mode. According to this, the information signal recorded when monitoring the technical condition of a product must contain a periodic component. The method is based on isolating this component and then approximating it with a model. The model of this component is described by a log-periodic function. The model parameters include the operating time of the product before repair and are determined in the periodic component approximation process by this model.