Comparative Analysis of Methods for Preparing Input Data for Neural Networks Used in Lightning Threat Assessment

The paper proposes automating the process of designing lightning protection systems. The impetus for improving the efficiency of this process came from the standardization procedures contained in IEC 62305, which are rigid and sometimes lead to overestimation or underestimation of the parameters of the lightning protection system. As a result of such actions, incorrect decisions may be made regarding applying a lightning protection system with suboptimally selected parameters for the objects of interest or even its complete omission. The automation of this process is based on two paths: starting with simulations of the electric field intensity distribution around objects at risk of lightning strikes and using neural networks designed to assess the locations particularly vulnerable to lightning threats. The process involving neural networks also includes a description of data preparation based on the Ansys Maxwell environment for use in an artificial neural network (ANN) based on Matlab and the Julia environment for use in an ANN based on Darknet.