Search

The use of kriging metamodels in simulation optimization has become increasingly popular during recent years. The majority of the algorithms so far uses the ordinary (deterministic) kriging approach for constructing the metamodel, assuming that solutions have been sampled with infinite precision. This is a major issue when the simulation problem is stochastic: ignoring the noise in the outcomes may not only lead to an inaccurate metamodel, but also to potential errors in identifying the optimal points among those sampled. Moreover, most algorithms so far have focused on single-objective problems. In this article, we test the performance of a multiobjective simulation optimization algorithm that contains two crucial elements: the search phase implements stochastic kriging to account for the inherent noise in the outputs when constructing the metamodel, and the accuracy phase uses a well-known multiobjective ranking and selection procedure in view of maximizing the probability of selecting the true Pareto-optimal points by allocating extra replications on competitive designs. We evaluate the impact of these elements on the search and identification effectiveness, for a set of test functions with different Pareto front geometries, and varying levels of heterogeneous noise. Our results show that the use of stochastic kriging is essential in improving the search efficiency; yet, the allocation procedure appears to lose effectiveness in settings with high noise. This emphasizes the need for further research on multiobjective ranking and selection methods. ispartof: European Journal of Operational Research vol:284 issue:1 pages:212-226 status: published

© 2018 Elsevier B.V. In this article, we consider the impact of finite production capacity on the optimal quality and pricing decisions of a make-to-stock manufacturer. Products are differentiated along a quality index; depending on the price and quality levels of the products offered, customers decide to either buy a given product, or not to buy at all. We show that, assuming fixed exogenous lead times and normally distributed product demands, the optimal solution has a simple structure (this is referred to as the load-independent system). Using numerical experiments, we show that with limited production capacity (which implies load-dependent lead times) the manufacturer may have an incentive to limit the quality offered to customers, and to decrease market coverage, especially in settings where higher product quality leads to higher congestion in production. Our findings reveal that the simple solution assuming load-independent lead times is suboptimal, resulting in a profit loss; yet, this profit loss can be mitigated by constraining the system utilization when deciding on quality and price levels. Our results highlight the importance of the relationship between marketing decisions and load-dependent production lead times. ispartof: EUROPEAN JOURNAL OF OPERATIONAL RESEARCH vol:272 issue:1 pages:195-206 status: published

In this article we investigate the unconstrained optimization (minimization) of the performance of a system that is modeled through a discrete-event simulation. In recent years, several algorithms have been proposed which extend the traditional Kriging-based simulation optimization algorithms (assuming deterministic outputs) to problems with noise. Our objective in this paper is to compare the relative performance of a number of these algorithms on a set of well-known analytical test functions, assuming different patterns of heterogeneous noise. We also apply the algorithms to a popular inventory test problem. The conclusions and insights obtained may serve as a useful guideline for researchers aiming to apply Kriging-based algorithms to solve engineering and/or business problems, and may be useful in the development of future algorithms.

This study seeks insights into the impact of inpatient boarding on emergency department (ED) congestion and capacity. To do so, we model the ED as a semi-open queueing network (SOQN) with limited resources (physicians and beds) and discontinuous patient service. We present a Markov-modulated fluid queue approach to efficiently calculate service levels, and show that boarding may cause the (expensive) physician resources to be starved, especially when the bed utilization is high. While the expected number of boarding patients has a primary impact on performance, we show that there is a secondary impact stemming from the expected boarding time and the boarding probability. Boarding reduction policies perform better if they focus on reducing expected boarding time instead of the decreasing probability of boarding. Our analysis and insights are applicable also to other SOQN settings where entities require more than one resource simultaneously (e.g., intensive care units, manufacturing systems, warehousing and transportation systems). ispartof: EUROPEAN JOURNAL OF OPERATIONAL RESEARCH vol:271 issue:3 pages:953-967 status: published

This article presents a novel heuristic for constrained optimization of computationally expensive random simulation models. One output is selected as objective to be minimized, while other outputs must satisfy given theshold values. Moreover, the simulation inputs must be integer and satisfy linear or nonlinear constraints. The heuristic combines (i) sequentialized experimental designs to specify the simulation input combinations, (ii) Kriging (or Gaussian process or spatial correlation modeling) to analyze the global simulation input/output data resulting from these designs, and (iii) integer nonlinear programming to estimate the optimal solution from the Kriging metamodels. The heuristic is applied to an (s, S) inventory system and a call-center simulation, and compared with the popular commercial heuristic OptQuest embedded in the Arena versions 11 and 12. In these two applications the novel heuristic outperforms OptQuest in terms of number of simulated input combinations and quality of the estimated optimum. ispartof: European Journal of Operational Research vol:202 issue:1 pages:164-174 status: published

In this paper we analyze how a nuclear magnetic resonance scanner can be managed more efficiently, simultaneously improving patient comfort (in terms of total time spent in the system) and increasing availability in case of emergency calls. By means of a superposition approach, all relevant data on the arrival and service process of different patient types are transformed into a general single server, single class queueing model. The objective function consists of the weighted average patient lead time, which is a multi-dimensional convex function of the different patient group sizes. The "optimal" patient group sizes are determined by means of a dedicated optimization routine. The model does not only provide a valuable aid for planning purposes, but also allows to model customer service. It is illustrated by means of real life data, obtained from the Virga Jesse Hospital (Hasselt, Belgium). (C) 2002 Elsevier B.V. All rights reserved. ispartof: European journal of operational research vol:151 issue:1 pages:181-192 status: published