: ..
:
: 114
:
: 2025
: .. // . - 2025. - . 114. - .41-64.
: , , ,
(.): parametric optimization, nonstationary processes, coefficient inverse problems, parabolic equations
: , , , , . , , , . , , , , . , , . , , . , , . , , , . . .
(.): When modeling nonstationary processes in continuous media by means of parabolic differential equations, we often encounter situations when the coefficient providing the connection between the left and right parts of the equation is described as a function of a set of variables, including the states of the medium under study. The recovery of this dependence, as a rule, requires the solution of inverse coefficient problems based on known states of the medium. In practice, this means that the inverse problem is solved relying, among other things, on some discrepancy between model data and known observations. Nevertheless, there are cases when such observations are critically small in time, for example, measurements of the state of the medium occur with a certain very large time step or only at the end of a nonstationary process. In such cases, retrospective observations contain time moments when the state of the medium is unknown, which makes it impossible to determine the error gradient for them and to restore the desired functional dependence with acceptable accuracy. In this paper, we propose an alternative view of the problem of restoring continuum coefficients for situations when the known states of the medium are much smaller than the unknown ones. A continuous nonstationary process was considered as a discrete process evolving in time, and a recurrent function of discrete state change was proposed. Based on this function, a numerical method for interpolating the error gradient between the expected and actual states of the medium within any two known states was proposed. The process of recovery of discrete values of coefficients at separate moments of time by means of the stochastic gradient descent method was demonstrated on the basis of a numerical model of a generalized parabolic equation with an arbitrary external influence on the boundary.
PDF
: 56, : 19, : 9.