https://repositorio.uai.cl//handle/20.500.12858/6746 Wed, 15 Jul 2026 11:23:15 GMT 2026-07-15T11:23:15Z https://repositorio.uai.cl//handle/20.500.12858/2570 Static risk-averse models with applications to mining Canessa, Gianpiero This thesis is centered around the development of methods and algorithms to solve different types of stochastic optimization problems that deal with risk. The first work in on chance-constrained problems (CCP), and the second on risk- averse two-stage stochastic problems (TSSP). The main challenge of stochastic Fri, 22 Oct 2021 00:00:00 GMT https://repositorio.uai.cl//handle/20.500.12858/2570 2021-10-22T00:00:00Z https://repositorio.uai.cl//handle/20.500.12858/2569 Mine Project Scheduling: Deterministic and Stochastic Models Lamas, Patricio En esta tesis, estudiamos la programación de proyectos con un foco en proyectos de minería. Primero, consideramos un problema determinístico simplificado y su correspondiente algoritmo de solución, el algoritmo de Lane. El objetivo de este algoritmo es optimizar la producción de una mina de rajo abierto de un solo metal y un solo procesador. Para esto, Lane propuso una política basada en el llamado “grado de corte” usado para determinar si el material extraído debería considerarse mineral o desperdicio. A pesar de que se pueden construir contraejemplos hipotéticos, nuestros experimentos computacionales muestran que el algoritmo de Lane produce la solución óptima a cada instancia real testeada, entregando un soporte sólido para su aplicación en la práctica. Luego, estudiamos un problema estocástico más realista cuya meta es determinar los tiempos de inicio de un conjunto de actividades con el objetivo de maximizar el valor presente neto del proyecto. Aquí asumimos 2 que cada actividad es caracterizada por una duración y beneficio aleatorios con una distribución de probabilidades conocida. Proponemos un enfoque proactivo/reactivo integrado que simultáneamente genera una ventana de tiempo inicial para el tiempo de inicio de cada actividad y una política de ejecución. Nuestros resultados computacionales muestran que, en general, las soluciones generadas por nuestro enfoque integrado dominan, en términos de valor presente neto y variabilidad de los tiempos de inicio de las actividades, aquellas generadas por las alternativas en la literatura. Finalmente, aplicamos exitosamente nuestro enfoque integrado a un proyecto minero subterráneo real. https://repositorio.uai.cl//handle/20.500.12858/2569 https://repositorio.uai.cl//handle/20.500.12858/2566 Applications of Integer Programming and Decomposition to Scheduling Problems: the Strategic Mine Planning Problem and the Bin Packing Problem with Time Lag Rivera, Orlando In scheduling problems, the goal is to assign time slots to a set of activities. In these problems, there are typically precedence constraints between activities that dictate the order in which they can be carried out and resource constraints that limit the number that can simultaneously be executed. In this thesis, we develop mixed integer programming methodologies, based on decomposition methods, for two very different classes of scheduling problems. These are the Strategic Open Pit Mine Planning Problem (SOPMP) and the Bin Packing Problem with Time Lags. Given a discretized representation of an orebody known as a block model, the SOPMP that we consider consists of defining which blocks to extract, when to extract them, and how or whether to process them, in such a way as to comply with operational constraints and maximize net present value. These problems are known to be very difficult due to the large size of real mine planning problems (eg, millions of blocks, dozens of years). They are also very important in the mining industry. Every major mining operation in the world must solve this problem, at the very least, on a yearly basis. In this thesis, we tackle the SOPMP in Chapters 2 and 3. In Chapter 2 we begin by studying a lagrangean algorithm developed by Dan Bienstock and Mark Zuckerberg (henceforth, the BZ algorithm) in 2009 for solving the LP relaxation of large instances of SOPMP. In this study we generalize the classes of problems that can be solved with the BZ algorithm, and show that it can be cast as a special type of column generation algorithm. We prove, for general classes of mixed integer programming problems, that the BZ relaxation provides a bound that lies between the LP relaxation and Dantzig- Wolfe bounds. We further develop computational speed-ups that improve the performance of the BZ algorithm in practice, and test these on a large collection of data-sets. In Chapter 3 we deal with the problem of computing integer-feasible solution to SOPMP. Using the BZ algorithm developed in Chapter 2, we develop heuristics for this. In addition, we develop pre-procesing algorithms that reduce problem size, and embed the BZ algorithm in a branch-and-cut framework that makes use of two new classes of cutting planes. When comparing the value of the heuristics to the LP relaxation bound, the average gap computed is close to 10%. However, when applying the pre-processing techniques and cutting planes, this is reduced to 1.5% at the root node. Four hours of branching further reduces this to 0.6%. In Chapter 4, the BPPTL is presented. This is a generalization of the Bin i Packing Problem in which bins must be assigned to time slots, while satisfying precedence constraints with lags. Two integer programming formulations are proposed: a compact formulation that models the problem exactly, and an extended formulation that models a relaxation. For two special cases of the problem, the case with unlimited bins per period and the case with one bin per period and non-negative time lags, we strengthen the extended formulation with a special family of constraints. We propose a branch-cut-and-price (BCP) algorithm to solve this formulation, with separation of integer and fractional solutions, and a strong diving heuristic. Computational experiments confirm that the BCP algorithm outperforms solving the compact formulation with a commercial solver. Using this approach we were able to optimally solve 70% of a class of previously open instances of this problem. Thu, 21 Oct 2021 00:00:00 GMT https://repositorio.uai.cl//handle/20.500.12858/2566 2021-10-21T00:00:00Z https://repositorio.uai.cl//handle/20.500.12858/2002 La importancia del espacio geográfico para minimizar el error de muestras representativas Truffello Robledo, Ricardo En el presente trabajo se discute la importancia del espacio geográfico en el contexto de la generación de marcos muestrales de encuestas, poniendo en tensión la premisa estadística tradicional de la aleatoriedad e independencia de las observaciones. Para esto se analiza el aporte de la geografía cuantitativa en la generación de metodologías de regionalización que permitan, de manera efectiva, mejorar el error muestral de las encuestas, enfocados principalmente en las áreas urbanas, en presencia de variables de estratificación con autocorrelación espacial. https://repositorio.uai.cl//handle/20.500.12858/2002