Methodology for constructing adaptive digital twins of technological objects based on hybrid models

The article presents a comprehensive technique for constructing adaptive digital twins of technological refining facilities based on the use of hybrid models combining rigorous thermodynamic calculations and iterative machine tuning procedures. The proposed approach combines the fundamental Peng–Robinson equation of state with van der Waals mixing rules and effective calibration of the model based on a limited sample covering only 10-12 technological modes. The scientific novelty of the study consists in three key aspects: firstly, an original algorithm for automatic parameter adjustment has been developed, ensuring an error in predicting the temperature profile of no more than 2.5%; secondly, a multi-criteria optimization procedure has been implemented with direct consideration of hydraulic constraints, such as the factor of contact device choking; thirdly, a flexible software architecture has been proposed, based on the integration of the Aspen HYSYS modeling environment and the Python scripting language via a COM interface. Practical testing of the technique on the K-1 industrial distillation column made it possible to achieve a reduction in specific energy consumption by 7.3%, which is equivalent to an annual economic effect of 4.52 million rubles. It is shown that the technique has the property of versatility and can be transferred to heat exchange, compressor and reactor equipment.