Forecasting Operating Parameters of a Technical Object Based on Random Process Description in the Form of Images on Curved Surfaces

To increase the reliability and safety of technical objects, it is necessary to diagnose and forecast their state. The article examines technical objects whose states depend on a set of operating parameters forming a time series system. Time periodicity is often observed in the process dynamics, but with significant and random differences between certain periods, which allows them to be represented as quasi-periodic random processes. Such processes are peculiar to many technical objects, for example, vibration of engines, turbines, compressors and hydraulic units. To increase forecast precision of the object state, we suggest to represent such processes as a spiral scan of a cylindrical image. This approach allows us to take into account mutual connections between quasi-periods and to apply image processing techniques to the analysis of quasi-periodic processes. When the process characteristics change smoothly, it is suggested to update model parameters according to an effective adaptive pseudo-gradient procedure. When the process properties change abruptly, an additional change in the model structure is provided. To assess the stability of an object and detect operational occurrences, fuzzy models are used. The effectiveness of the proposed methods and approaches is demonstrated on the example of assessing a hydraulic unit state.