Francisco Trespalacios

An MILP-MINLP decomposition method for the global optimization of a source based model of the multiperiod blending problem

Abstract The multiperiod blending problem involves binary variables and bilinear terms, yielding a nonconvex MINLP. In this work we present two major contributions for the global solution of the problem. The first one is an alternative formulation of the problem. This formulation makes use of redundant constraints that improve the MILP relaxation of the MINLP. The second contribution is an algorithm that decomposes the MINLP model into two levels. The first level, or master problem, is an MILP relaxation of the original MINLP. The second level, or subproblem, is a smaller MINLP in which some of the binary variables of the original problem are fixed. The results show that the new formulation can be solved faster than alternative models, and that the decomposition method can solve the problems faster than state of the art general purpose solvers.

Improved Big-M reformulation for generalized disjunctive programs

Abstract In this work, we present a new Big-M reformulation for Generalized Disjunctive Programs. Unlike the traditional Big-M reformulation that uses one M-parameter for each constraint, the new approach uses multiple M-parameters for each constraint. Each of these M-parameters is associated with each alternative in the disjunction to which the constraint belongs. In this way, the proposed MINLP reformulation is at least as tight as the traditional Big-M, and it does not require additional variables or constraints. We present the new Big-M, and analyze the strength in its continuous relaxation compared to that of the traditional Big-M. The new formulation is tested by solving several instances with an NLP-based branch and bound method. The results show that, in most cases, the new reformulation requires fewer nodes and less time to find the optimal solution.

Algorithmic Approach for Improved Mixed-Integer Reformulations of Convex Generalized Disjunctive Programs

In this work, we propose an algorithmic approach to improve mixed-integer models that are originally formulated as convex generalized disjunctive programs (GDPs). The algorithm seeks to obtain an improved continuous relaxation of the mixed-integer linear and mixed-integer nonlinear programming (MILP/MINLP) model reformulation of the GDP while limiting the growth in the problem size. There are three main stages that form the basis of the algorithm. The first one is a presolve, consequence of the logic nature of GDP, which allows us to reduce the problem size, find good relaxation bounds, and identify properties that help us determine where to apply a basic step. The second stage is the iterative application of basic steps, selecting where to apply them and monitoring the improvement of the formulation. Finally, we use a hybrid reformulation of GDP that seeks to exploit both of the advantages attributed to the two common GDP-to-MILP/MINLP transformations, the Big-M, and the Hull reformulation. We illustrate the application of this algorithm with several examples. The results show the improvement in the problem formulations by generating models with improved relaxed solutions and relatively small growth of continuous variables and constraints. The algorithm generally leads to reduction in the solution times.

Cutting Plane Algorithm for Convex Generalized Disjunctive Programs

In this work, we propose a cutting plane algorithm to improve optimization models that are originally formulated as convex generalized disjunctive programs. Generalized disjunctive programs are traditionally reformulated as mixed-integer nonlinear programming (MINLP) problems using either the big M (BM) or the hull reformulation (HR). The former yields a smaller MILP/MINLP problem, whereas the latter yields a tighter one. The HR can be further strengthened by using the concept of basic step from disjunctive programming. The proposed algorithm uses the strengthened formulation to derive cuts for the big-M formulation, generating a stronger formulation with small growth in problem size. We test the algorithm with several instances. The results show that the algorithm improves generalized disjunctive programming convex models, in the sense of providing formulations with stronger continuous relaxations than the BM formulation, with few additional constraints. In general, the algorithm also leads to a reduction in the solution time of the problems.

Symmetry breaking for generalized disjunctive programming formulation of the strip packing problem

In this work we present a new generalized disjunctive programming (GDP) formulation for the strip packing problem. The new formulation helps to break some of the symmetry that arises in this problem. The new formulation is further improved for the case in which the heights and lengths of the rectangles are integer numbers. The GDP model can be formulated and solved as a mixed-integer linear programming (MILP) model, using different GDP-to-MILP reformulations. The results show that the MILP reformulations of the new GDP model (and its improvement for rectangles with integer heights and widths) can be solved faster than the previously proposed GDP formulation.

Recent Progress Using Matheuristics for Strategic Maritime Inventory Routing

This chapter presents an extensive computational study of simple, but prominent matheuristics (i.e., heuristics that rely on mathematical programming models) to find high quality ship schedules and inventory policies for a class of maritime inventory routing problems. Our computational experiments are performed on a test bed of the publicly available MIRPLib instances. This class of inventory routing problems has few constraints relative to some operational problems, but is complicated by long planning horizons. We compare several variants of rolling horizon heuristics, K-opt heuristics, local branching, solution polishing, and hybrids thereof. Many of these matheuristics substantially outperform the commercial mixed-integer programming solvers CPLEX 12.6.2 and Gurobi 6.5 in their ability to quickly find high quality solutions. New best known incumbents are found for 26 out of 70 yet-to-be-proved-optimal instances and new best known bounds on 56 instances.

Lagrangean relaxation of the hull-reformulation of linear generalized disjunctive programs and its use in disjunctive branch and bound

In this work, we present a Lagrangean relaxation of the hull-reformulation of discrete-continuous optimization problems formulated as linear generalized disjunctive programs (GDP). The proposed Lagrangean relaxation has three important properties. The first property is that it can be applied to any linear GDP. The second property is that the solution to its continuous relaxation always yields 0–1 values for the binary variables of the hull-reformulation. Finally, it is simpler to solve than the continuous relaxation of the hull-reformulation. The proposed Lagrangean relaxation can be used in different GDP solution methods. In this work, we explore its use as primal heuristic to find feasible solutions in a disjunctive branch and bound algorithm. The modified disjunctive branch and bound is tested with several instances with up to 300 variables. The results show that the proposed disjunctive branch and bound performs faster than other versions of the algorithm that do not include this primal heuristic.

Cutting planes for improved global logic-based outer-approximation for the synthesis of process networks

Abstract In this work, we present an improved global logic-based outer-approximation method (GLBOA) for the solution of nonconvex generalized disjunctive programs (GDP). The GLBOA allows the solution of nonconvex GDP models, and is particularly useful for optimizing the synthesis of process networks, which yields MINLP models that can be highly nonconvex. However, in many cases the NLP that results from fixing the discrete decisions is much simpler to solve than the original problem. The proposed method exploits this property. Two enhancements to the basic GLBOA are presented. The first enhancement seeks to obtain feasible solutions faster by dividing the basic algorithm into two stages. The first stage seeks to find feasible solutions faster by restricting the solution time of the problems and diversifying the search. The second stage guarantees the convergence by solving the original algorithm. The second enhancement seeks to tighten the lower bound of the algorithm by the use of cutting planes. The proposed method for obtaining cutting planes, the main contribution of this work, is a separation problem based on the convex hull of the feasible region of a subset of the constraints. Results and comparison with other global solvers show that the enhancements improve the performance of the algorithm, and that it is more effective in the tested problems at finding near optimal solutions compared to general-purpose global solvers.

Global optimization algorithm for capacitated multi-facility continuous location-allocation problems

In this paper we propose a nonlinear Generalized Disjunctive Programming model to optimize the 2-dimensional continuous location and allocation of the potential facilities based on their maximum capacity and the given coordinates of the suppliers and customers. The model belongs to the class of Capacitated Multi-facility Weber Problem. We propose a bilevel decomposition algorithm that iteratively solves a discretized MILP version of the model, and its nonconvex NLP for a fixed selection of discrete variables. Based on the bounding properties of the subproblems,