Surrogate models of electrophysical characteristics of glass parts of high-voltage insulators

The article considers the development of a surrogate model as an element of a computer vision system to identify internal defects in insulating parts of complex shape. It determines the principles of optimal construction of a physical model necessary to obtain simulation data in order to analyze the electrophysical characteristics of high-voltage insulators with defects. It also defines a set of functional dependencies for the analysis of the electrical strength of the sample and the principles of selecting arguments for these dependencies. Using the example of a glass part of the PS-70E insulator, it shows that a numerical model in the form of a polynomial of five seventh-order variables, taking into account the change in the shape of elliptical gas inclusions and their distribution, is sufficient to analyze the electrical strength of the insulator. The relative error in determining the main characteristics of the electric field in the defect based on the results of simulation modeling and calculation using a surrogate model is no more than 10 %. The use of a surrogate model helps to identify the areas where the maximum electric field strength will be observed in defects, which makes it possible for the computer vision system to assess the potential danger of defects. Using the polynomial form of the surrogate model simplifies the process of creating a model, since the construction algorithm is deterministic. This advantage can be used in the construction of surrogate models of insulation structures, since their physical properties are sufficiently homogeneous throughout the entire product and they can be approximated by a polynomial function of a small degree. Besides, the advantage of the polynomial model was determined by the relative simplicity of obtaining simulation data.