Clifford A. Meyer

Global optimization in the 21st century: Advances and challenges

This paper presents an overview of the research progress in global optimization during the last 5 years (1998–2003), and a brief account of our recent research contributions. The review part covers the areas of (a) twice continuously differentiable nonlinear optimization, (b) mixedinteger nonlinear optimization, (c) optimization with differential-algebraic models, (d) optimization with grey-box/black-box/nonfactorabl e models, and (e) bilevel nonlinear optimization. Our research contributions part focuses on (i) improved convex underestimation approaches that include convex envelope results for multilinear functions, convex relaxation results for trigonometric functions, and a piecewise quadratic convex underestimator for twice continuously differentiable functions, and (ii) the recently proposed novel generalized BB framework. Computational studies will illustrate the potential of these advances.

Trilinear Monomials with Mixed Sign Domains: Facets of the Convex and Concave Envelopes

Convex underestimators of nonconvex functions, frequently used in deterministic global optimization algorithms, strongly influence their rate of convergence and computational efficiency. A good convex underestimator is as tight as possible and introduces a minimal number of new variables and constraints. Multilinear monomials over a coordinate aligned hyper-rectangular domain are known to have polyhedral convex envelopes which may be represented by a finite number of facet inducing inequalities. This paper describes explicit expressions defining the facets of the convex and concave envelopes of trilinear monomials over a box domain with bounds of opposite signs for at least one variable. It is shown that the previously used approximations based on the recursive use of the bilinear construction rarely yield the convex envelope itself.

Convex envelopes for edge-concave functions

Abstract.Deterministic global optimization algorithms frequently rely on the convex underestimation of nonconvex functions. In this paper we describe the structure of the polyhedral convex envelopes of edge-concave functions over polyhedral domains using geometric arguments. An algorithm for computing the facets of the convex envelope over hyperrectangles in ℝ3 is described. Sufficient conditions are described under which the convex envelope of a sum of edge-concave functions may be shown to be equivalent to the sum of the convex envelopes of these functions.

Trilinear Monomials with Positive or Negative Domains: Facets of the Convex and Concave Envelopes

Approximations of the convex envelope of nonconvex functions play a central role in deterministic global optimization algorithms and the efficiency of these algorithms is highly infuenced by the tightness of these approximations. McCormick (1976), and AlKhayyal and Falk (1983) have shown how to construct the convex envelope of individual bilinear terms over a rectangular domain. Rikun (1997) has shown that the convex hull of multilinear monomials over a rectangular domain is polyhedral. Approximations of the convex envelope for higher order multilinear terms have been based on the recursive use of this bilinear construction. Only under very special circumstances, however, do these approximations yield the convex envelope itself. Explicit expressions defining the facets of the convex and concave envelopes for trilinear monomials, with positive or negative bounded domains for each variable, are derived in this paper.

Convex Underestimation of Twice Continuously Differentiable Functions by Piecewise Quadratic Perturbation: Spline αBB Underestimators

This paper describes the construction of convex underestimators for twice continuously differentiable functions over box domains through piecewise quadratic perturbation functions. A refinement of the classical α BB convex underestimator, the underestimators derived through this approach may be significantly tighter than the classical αBB underestimator. The convex underestimator is the difference of the nonconvex function f and a smooth, piecewise quadratic, perturbation function, q. The convexity of the underestimator is guaranteed through an analysis of the eigenvalues of the Hessian of f over all subdomains of a partition of the original box domain. Smoothness properties of the piecewise quadratic perturbation function are derived in a manner analogous to that of spline construction.

Generalized Geometric Programming Problems

In this chapter, we will discuss test problems that arise from generalized geometric programming applications. For a thorough theoretical and algorithmic exposition of global optimization approaches for generalized geometric programming problems, the reader is directed to the article of Maranas and Floudas (1997), and the book by Floudas (2000).

Quadratic Programming Problems

In this chapter nonconvex quadratic programming test problems are considered. These test problems have a quadratic objective function and linear constraints. Quadratic programming has numerous applications (Pardalos and Rosen (1987), Floudas and Visweswaran (1995)) and plays an important role in many nonlinear programming methods. Recent methods of generating challenging quadratic programming test problems and disjointly constrained bilinear programming test problems can be found in the work of Vicente et al. (1992) and Calamai et al. (1993). Furthermore, a very broad class of difficult combinatorial optimization problems such as integer programming, quadratic assignment, and the maximum clique problem can be formulated as nonconvex quadratic programming problems.

Twice Continuously Differentiable NLP Problems

Twice continuously differentiable NLPs represent a very broad class of problems with diverse applications in the fields of engineering, science, finance and economics. Specific problems include phase equilibrium characterization, minimum potential energy conformation of clusters and molecules, distillation sequencing, reactor network design, batch process design, VLSI chip design, protein folding, and portfolio optimization.

Nonlinear Systems of Equations

The solution of nonlinear systems of equations is an important problem in engineering. Applications include identifying the multiple steady states of a reactor network (Folger, 1986; Schlosser and Feinberg, 1994), predicting the azeotropes formed in a nonideal mixture (Fidkowski et al., 1993; Maranas et al., 1996), finding the steady-states of a flowsheet or part of a flowsheet (Zhang, 1987; Bullard and Biegler, 1991; Wilhelm and Swaney, 1994; Bekiaris et al., 1993), identifying equilibrium points in multiphase systems (Heidemann and Mandhane, 1973; Seader et al., 1990) .

Semidefinite Programming Problems

Semidefinite programming involves the minimization of a linear function subject to the constraint that an affine combination of symmetric matrices is positive semidefinite. Several types of problems can be transformed to this form. This constraint is in general nonlinear and nonsmooth yet convex. Semidefinite programming can be viewed as an extension of linear programming and reduces to the linear programming case when the symmetric matrices are diagonal.