Developing a Technique for Constructing Surrogate Models of Equipment on the Example of a Hydraulic Model of a Gas-Removal Device

• Published in: Thermal engineering 2023-02, Vol.70 (2), p.139-144
• Main Authors: Konovalov, I. A., Sokolov, A. N., Barinov, A. A., Zyryanova, T. K.
• Format: Article
• Language: English
• Subjects: Algorithms; Approximation; Complexity; Computer simulation; Engineering; Engineering Thermodynamics; Heat and Mass Transfer; Hydraulic equipment; Hydraulic models; Hydraulics; Mathematical models; Multiple objective analysis; Nuclear Power Plants; Nuclear safety; Optimization; Parameters; Power management; Training
• ISSN: 0040-6015; 1555-6301
• DOI: 10.1134/S0040601523020040

At present, the power-engineering industry is faced with the urgent task of increasing the efficiency and safety of the equipment being designed. Due to the growing complexity of the developed units, an increasingly detailed analysis of thermal-hydraulic processes is required (including the use of three-dimensional CFD codes), which leads to an increase in the time spent on simulation. At the same time, increasing the efficiency of complex thermal-hydraulic systems with many variable technical parameters that affect the performance of the system can be justified using mathematical algorithms for multiobjective optimization, which require a large number of variant calculations (or experiments) of the system under consideration. In this case, the use of CFD codes for modeling individual geometrically complex elements in the optimization cycle becomes difficult since it leads to unacceptable time costs. This problem can be solved using the approximation dependence (surrogate model) of the input and output parameters, built on the basis of some training sample obtained during the calculation or experiment, instead of the original complex computational mathematical model. However, in this case, with an increase in the dimension of the problem, an increase in the training set is required. In the framework of this work, a method is proposed for representing a surrogate model of high dimension as a set of approximation dependencies with a smaller number of dimensions, which allows training more accurate models on smaller samples. The methodology for generating training samples is considered, and the methodology for constructing hydraulic surrogate models of power equipment is presented using the example of a gas-removal device for a low-power nuclear power plant. A good agreement between the constructed surrogate model and the results of numerical simulation is obtained.