Том 57, № 4 (2017)

The Appell function F₁ (i.e., a generalized hypergeometric function of two complex variables) and a corresponding system of partial differential equations are considered in the logarithmic case when the parameters of F₁ are related in a special way. Formulas for the analytic continuation of F₁ beyond the unit bicircle are constructed in which F₁ is determined by a double hypergeometric series. For the indicated system of equations, a collection of canonical solutions are presented that are two-dimensional analogues of Kummer solutions well known in the theory of the classical Gauss hypergeometric equation. In the logarithmic case, the canonical solutions are written as generalized hypergeometric series of new form. The continuation formulas are derived using representations of F₁ in the form of Barnes contour integrals. The resulting formulas make it possible to efficiently calculate the Appell function in the entire range of its variables. The results of this work find a number of applications, including the problem of parameters of the Schwarz–Christoffel integral.

For the wave equation, time-optimal control problems with two-sided boundary controls of three basic types are considered in classes of weak generalized solutions. Stable algorithms for computing approximate optimal times and optimal boundary controls are developed by modifying algorithms proposed earlier for the case of strong generalized solutions. The approximate solutions are proved to converge as the parameters of the finite-dimensional approximation are asymptotically refined and the errors in the specified terminal functions are reduced.

The problem of optimizing loading places and corresponding load response functions with respect to objects described by systems of loaded ordinary differential equations is solved numerically. Analytical formulas for the gradient of the functional with respect to the optimized load parameters are derived to solve the problem by applying first-order numerical methods. Results of numerical experiments are presented. The approach proposed can also be used to optimize load parameters in distributed systems described by partial differential equations, which are reduced to the considered problem by applying the method of lines.

A method for constructing pseudo-holomorphic solutions to strongly nonlinear singularly perturbed systems of differential equations, which a logical continuation of the Lomov regularization method, is proposed. The existence of integrals of such systems, holomorphic in the small parameter, is proven, and sufficient conditions for the convergence of expansion of solutions to these systems in powers of the small parameter in the usual sense are obtained.

In this paper, 1-exact vertex-centered finite-volume schemes with an edge-based approximation of fluxes are constructed for numerically solving hyperbolic problems on hybrid unstructured meshes. The 1-exactness property is ensured by introducing a new type of control volumes, which are called semitransparent cells. The features of a parallel algorithm implementing the computations using semitransparent cells on modern supercomputers are described. The results of solving linear and nonlinear test problems are given.

The multidimensional quasi-gasdynamic system written in the form of mass, momentum, and total energy balance equations for a perfect polytropic gas with allowance for a body force and a heat source is considered. A new conservative symmetric spatial discretization of these equations on a nonuniform rectangular grid is constructed (with the basic unknown functions—density, velocity, and temperature—defined on a common grid and with fluxes and viscous stresses defined on staggered grids). Primary attention is given to the analysis of entropy behavior: the discretization is specially constructed so that the total entropy does not decrease. This is achieved via a substantial revision of the standard discretization and applying numerous original features. A simplification of the constructed discretization serves as a conservative discretization with nondecreasing total entropy for the simpler quasi-hydrodynamic system of equations. In the absence of regularizing terms, the results also hold for the Navier–Stokes equations of a viscous compressible heat-conducting gas.

This paper describes a new efficient conjugate subgradient algorithm which minimizes a convex function containing a least squares fidelity term and an absolute value regularization term. This method is successfully applied to the inversion of ill-conditioned linear problems, in particular for computed tomography with the dictionary learning method. A comparison with other state-of-art methods shows a significant reduction of the number of iterations, which makes this algorithm appealing for practical use.