An algorithm and software implementation of test object model synthesis based on the solution of the nonparametric identification equation

The paper considers the development of the theory of testing in general and the experimental-theoretical method in particular. In the aspect of this issue, the authors have developed an algorithm for synthesizing a model of a test object based on solving the equation of nonparametric identification of a dynamic system using hyperdelta approximation and the Laplace transform. Unlike the existing ones, the algorithm is applicable to input and output signals of arbitrary shape and physical quantities. In addition, it does not require large computing resources. Taking into account these features, the algorithm enables formalizing a multidimensional relationship between factors and performance characteristics of the test object through repeated use for different input and output signals. The authors have implemented a mathematical library for identifying a test object model and an application with a graphical user interface for automating calculations using the C++ and Python programming languages. The presented software solution is made similar to classical machine learning models. To substantiate the possibility of using the developed algorithm, the authors carried out a computational experiment that involved various types of input and output signals (periodic, non-periodic and random) with different hyperdelta approximation accuracy. Based on the results of the computational experiment, the authors have made recommendations on using the algorithm. In particular, they recommended to increase the number of initial moments of the hyperdelta approximation at high amplitudes of the output signal.