Industrial Automatic Control Systems and Controllers Annotation << Back An optimization technique for the plants governed by a system of nonlinear algebraic equations or inequalities Scherbashin Yu.D, Sidorenko V.Ya. An approximating programming technique with gradually increasing/decreasing basis rank is discussed. The paper shows that if a solution is delivered in the vertex of polyhedron – on the border of the intersection of n bounding hypersurfaces (where n is the rank of space of decision variables), then the basis rank equals n; if the solution is delivered on limiting polyhedron’s facets or edges, then the rank of the basis decreases; if the solution is delivered in the interior of the accessible region, then the rank of the basis equals zero, and the change of X at the last steps matches the steepest descent/ascent algorithm. The second feature of the new technique is the application of quadratic approximation of discrepancy change ??i(Х) along the admissible direction – the ray ? – of the linear combination of the current basic cone’s edges. The quadratic approximation enables step size increase as against the simplest approximating programming techniques. E-mail: nedzelsky@zfs.lg.ua