Informs 2015

456 / 552

INFORMS Philadelphia – 2015
454
3 - Optimizing Military Medevac Dispatching:
A Multi-objective Markov Decision Process Model
Benjamin Grannan, Virginia Military Institute, Lexington, VA,
United States of America, , Nathaniel
Bastian, Lawrence Fulton, Mort Webster, Paul Griffin
We present an infinite horizon Multi-Objective Markov Decision Process model to
optimize sequential resource allocation decision-making in the military medical
evacuation of wartime casualties, which consists of identifying how many/which
air assets to dispatch in response to a casualty event and which mobile hospital to
transport the patients to. These sequential decisions are complicated due to
uncertain casualty demand, distinguishable mobile hospitals, and multiple
conflicting objectives.
4 - Stochastic Multi-objective Auto-optimization for Resource
Allocation Decision-making
Nathaniel Bastian, Pennsylvania State Univeristy, 362 Leonhard
Building, University Park, PA, 16823, United States of America,
, Lawrence Fulton, Benjamin
Grannan, Tahir Ekin, Hyojung Kang, Paul Griffin
The military health system is a large, centrally-funded and controlled health
system that is challenged to provide healthcare delivery and health services at
certain quality and workload levels with a fixed amount of input resources. We
present a stochastic multi-objective auto-optimization model to help health
system decision-makers automatically re-allocate input resources across hospitals
for difference levels of resource uncertainty as to optimize overall system
performance.
WD02
02-Room 302, Marriott
Scheduling IV
Contributed Session
Chair: Hossein Soroush, Professor, Kuwait University, Dept. of Stat. &
Opns. Res., POB 5969, Safat, 13060, Kuwait,
1 - Heuristics for Scheduling Parallel Machines to Minimize Weighted
Squared Tardiness
Jeffrey Schaller, Professor, Eastern Connecticut State University,
83 Windham St., Department of Business Administration,
Willimantic, CT, 06226, United States of America,
Jorge Valente
This paper considers a problem in which there is a set of jobs to be sequenced on
identical parallel machines. Each job has a weight and the objective is to sequence
the jobs to minimize total weighted squared tardiness. Several heuristics are
presented and are tested on randomly generated problems of various numbers of
jobs, numbers of machines, due date tightness and due date ranges. The results
show that some heuristics finds good solutions in a minimal amount of processing
time.
2 - Investigation of Algorithms for No-Wait Flowshops
Ali Allahverdi, Professor, Kuwait University, P.O. Box 5969,
Safat, Kuwait,
The m-machine no-wait flowshop scheduling problem of minimizing total
completion time is addressed where makespan is constrained to be less than a
certain value and where setup times are considered as separate from processing
times. Several new algorithms are proposed and shown to be efficient.
3 - Synchronized Scheduling in Hybrid Flowshop with Dynamic
Customer Order Arrivals
Jian Chen, The University of Hong Kong, LG 108,
Composite Building, Hong Kong, Hong Kong - PRC,
George Q. Huang
We consider a synchronized scheduling of dynamic arriving customer orders in a
hybrid flowshop. The objective is to achieve synchronized production that all
products of a same order are simultaneously completed so that they are shipped
in a batch. We decompose the dynamic problem into a series of periodic sub-
problems. A mixed integer programming model is formulated for sub-problem
which is solved by genetic algorithm. Numerical studies obtain several significant
findings.
4 - Simulation Based Approximate Dynamic Programming for
Stochastic Resource-constrained Project Scheduling
Yasin Gocgun, Assistant Professor, Istanbul Kemerburgaz
University, Mahmutbey Mah., Dilmenler 26, Istanbul,
Turkey, , Archis Ghate,
Mahshid Parizi
We study dynamic stochastic resource-constrained project scheduling problems.
We formulate these problems as Markov decision processes (MDPs). Since the
state and actions spaces of the underlying MDP grow exponentially in parameters
such as the number of project types and queue capacity, we pursue an
approximate dynamic programming (ADP) approach. We compare the
performance of the ADP technique against the myopic method.
5 - Scheduling with Convex Resource Allocation, Learning Effects,
Job Deterioration, and Setup Times
Hossein Soroush, Professor, Kuwait University,
Dept. of Stat. & Opns. Res., POB 5969, Safat, 13060, Kuwait,
We study single machine scheduling problems with convex resource dependent
processing times, job learning and deterioration, and setup times. Polynomial time
algorithms are proposed to find the optimal sequences and the optimal resource
allocations that either minimize a composite function of some criteria subject to
the constraint that the total resource cost does not exceed a given amount, or
minimize the total resource cost such that the composite function does not
surpass a specific limit.
WD03
03-Room 303, Marriott
Inventory Management for Supply Chains II
Contributed Session
Chair: Yang Bo, PhD, University of Texas at Dallas,
800 West Campbel Rd., Richardson, TX, United States of America,
1 - A Three-Level Supply Chain with Up and Down-Stream Trade
Credit Periods Linked to Ordered Quantity
Roshanak Akram, university of Tennessee, Knoxville, 727 Front
Avenue, Apartment 9, Knoxville, TN, 37902, United States of
America, Rupy Sawhney
An EPQ model from manufacturer prospective in a 3-echelon supply chain is
developed to minimize total cost considering that a trade credit is offered to the
manufacturer, if his order quantity is more than a predetermined value.
Customers are permitted to pay off manufacturer in a same or contrary credit
period. Defective items can be produced and all would become perfect after
rework. Optimal solutions for production lot size and backlogged shortage are
obtained based on convex optimization.
2 - A Note on Integrality in Deterministic and Stochastic
Inventory Models
Yang Bo, PhD, University of Texas at Dallas, 800 West Campbell
Rd., Richardson, TX, United States of America,
The “integrality” question for dynamic optimization models of inventory control
asks if there exists an integral optimal policy, given integral initial inventory
levels, capacities and demand realizations. For single-product, stochastic-demand
problems in multi-echelon assembly systems, the answer is yes (Chen et al.
2013). For single-product, multi-echelon distribution and assembly-distribution
systems, integrality holds under deterministic demands, but fails to hold under
stochastic demands.
3 - A Joint Replenishment Problem with Dissimlar Items
Linda Li, Student, University of Alabama, 1105 17th Street,
5102A, Tuscaloosa, AL, 35401, United States of America,
, Charles Schmidt
We study a joint replenishment problem with dissimilar items and demand of
Poisson process. A class of continuous review policies is considered. The class is
defined by three characteristics. Finding the optimal policy out of the class is
demonstrated to be a combinatorial optimization problem. The findings reveal
that the joint inventory position (a vector) is actually a subordinated continuous
time Markov Chain. A new methodology is developed to solve this type of model.
WD04
04-Room 304, Marriott
Business Applications II
Contributed Session
Chair: James Ang, Research Advisor, The Logistics Institute Asia Pacific,
National University of Singapore, #04-01 21 Heng Mui Keng Terrace,
Singapore, 119613, Singapore,
1 - A Framework for Customer Analytics on Massive Data Sets
Ming Xie, IBM, Diamond Bld, ZGC Software Park, Beijing, China,
Changrui Ren, Tianzhi Zhao, Miao He,
Jian Xu, Zhen Huang, Yuhui Fu
With the rapid growth of the amount of data accumulated, companies are facing
opportunities and challenges for efficiently analyzing these massive data to
generate insights and help business. Here we present a framework that leverages
big data foundation to provide efficient customer analytical models designed to
support the key customer service scenarios in banks. This framework has been
verified in real cases.
WD02