Том 56, № 8 (2016)

An algorithm for computing the covering constant for the restriction of a linear operator to a cone defined by a finite set of inequalities is proposed. After a finite number of steps, the algorithm reduces the original problem to one of finding the eigenvalues of linear operators.

Adaptive methods for the polyhedral approximation of the convex Edgeworth–Pareto hull in multiobjective monotone integer optimization problems are proposed and studied. For these methods, theoretical convergence rate estimates with respect to the number of vertices are obtained. The estimates coincide in order with those for filling and augmentation H-methods intended for the approximation of nonsmooth convex compact bodies.

For a singularly perturbed elliptic boundary value problem, an asymptotic expansion of the boundary-layer solution is constructed and justified in the case when the boundary layer consists of three zones with different behavior of the solution, which is caused by the multiplicity of the root of the degenerate equation.

The hodograph method is used to construct a solution describing the interaction of weak discontinuities (rarefaction waves) for the problem of mass transfer by an electric field (zonal electrophoresis). Mathematically, the problem is reduced to the study of a system of two first-order quasilinear hyperbolic partial differential equations with data on characteristics (Goursat problem). The solution is constructed analytically in the form of implicit relations. An efficient numerical algorithm is described that reduces the system of quasilinear partial differential equations to ordinary differential equations. For the zonal electrophoresis equations, the Riemann problem with initial discontinuities specified at two different spatial points is completely solved.

A modified splitting method for solving the nonstationary kinetic equation of particle (neutron) transport without iteration with respect to the collision integral is proposed. According to the modification, the solutions of the first-stage integrodifferential equations and the collision integrals are found using analytical rather than finite-difference methods. The solution method is naturally extended to multidimensional problems and is well suited for massive parallelism.

A numerical method for the integration of three-dimensional Navier–Stokes equations for compressible fluid as applied to direct numerical simulation is proposed. By way of example, the boundary layer on a plate is simulated. The computations were carried out for Re_θ = 1500. The computational grid consisted of a half billion nodes. The flow region includes the laminar, transitional, and turbulent zones. The numerically obtained distributions of average velocity, friction, and pulsations are compared with experimental data and available numerical solutions.

A number of transformations are introduced that are invariant under minimization problems and make it possible to reduce the maximum possible number of distinct columns in the matrix of zeros of an arbitrary binary function of multivalued arguments. As a result, simpler disjunctive normal forms are constructed. Complexity bounds for the constructed disjunctive normal forms of arbitrary binary functions of k-valued arguments are given.