Informs 2015

476 / 552

INFORMS Philadelphia – 2015
479
2 - Addressing Complexity of Hurricane Sandy with Innovative
Kingdon’s Model
Eliot Evans, Lt Col & Graduate Student, George Mason University
School of Public Policy, Government, & International Affairs,
12308 Cicero Drive, Alpharetta, GA, 30022,
United States of America,
Losses due to the impact of Hurricane Sandy in 2012 raise the concern of the
effectiveness of disaster management and its operations. FEMA’s Hurricane Sandy
After-Action Report revealed three significant problems 1) lack of collaboration 2)
inadequate survivors’ needs met, 3) shortage of an agile, professional emergency
management workforce. This research aims to analyze the complexity of
Hurricane Sandy and its problems, to propose an agenda and alternatives, and to
recommend public policies.
3 - throughput Analysis of Reserve Component Mobilization T
raining Capacity
Katharina Best, Associate Operations Researcher, The RAND
Corporation, 1200 S. Hayes St, Arlington, VA, 22202,
United States of America, Jeremy Eckhause, Igor
Mikolic-torreira, Michael Linick
Army Reserve Component units require administrative processing and varying
amounts of high-quality training at specialized installations before deploying to
contingency locations. Capacity at such facilities is limited and opening bases
quickly can be problematic. We present a mixed-integer programming model that
optimizes training schedules under different assumptions about training time,
facilities ramp-up, unit type prioritization, demand timing, and Active/Reserve
Component force mix.
4 - Optimal Multi-stage Allocation via Approximate
Dynamic Programming
Darryl Ahner, Asst Professor, Air Force Institute of Technology,
2950 Hobson Way, Wright-Patterson AFB, OH, 45433-7765,
United States of America, , Carl Parson
We consider the optimal allocation of resources over multiple stages to a
collection of tasks with the objective of maximizing the reward for completing
tasks where the task arrivals follow a known distribution, namely stochastic
weapon-target assignment. Simulation and mathematical programming are used
within a dynamic programming framework to update functional approximations
representing future rewards using subgradient information and thereby
determine allocation strategies.
WE02
02-Room 302, Marriott
Scheduling V
Contributed Session
Chair: Majid Algwaiz, Engineering Specialist, Saudi Aramco Oil
Company, P.O. Box 19422, Dhahran, 31311, Saudi Arabia,
1 - A Genetic Algorithm for the Resource Leveling Problem with
Generalized Precedence Relations
Hongbo Li, Shanghai University, School of Management, Shangda
Road 99, Shanghai, 200444, China, ,
Yinbin Liu, Li Xiong
We present a bi-chromosome based genetic algorithm (BGA) for the resource
leveling problem with generalized precedence relations. In the BGA, a solution is
represented by a bi-chromosome that consists of two parts: a random key vector
and a percentage based shift vector. To demonstrate the effectiveness of our BGA,
we conduct extensive computational experiments on a set of benchmarks with up
to 500 activities and compare the BGA with two best metaheuristics in the
literature.
2 - Results on throughput Maximization with Limited
Advance Information
Ishwar Murthy, Professor, Indian Institute of Management
Bangalore, Bannerghatta Road, Bangalore, 560076, India,
We consider throughput maximization, given limited advance information. For
problems where job lengths are equal, we propose an on-line algorithm whose
competitive ratio improves as the duration of the advance information increases.
Further, the performance of this on-line algorithm asymptotically approaches that
of the off-line algorithm. More importantly, we help to identify the structure of
the worst case instances – those that correspond to the competitive ratios.
3 - A Lower Bound Analysis for the Flowshop Scheduling Problem
with Makespan Minimization
Carlos Ernani Fries, Professor, Federal University of Santa
Catarina, Caixa Postal 5185, Florianopolis, SC, 88040-970, Brazil,
, Bruno De Souza Alves
This paper deals with a lower bound (LB) analysis for makespan measure of FSP.
The LB measures are compared with solutions obtained with exact models and
the popular CDS heuristic. Simulations varying the number of jobs, machines and
processing times show that solutions discrepancies tend to increase until N less
than M and decrease for N greater than M, with largest discrepancy observed for
N equal M. The divergences tend to be larger when greater variability on
processing times is considered.
4 - Stochastic Patient Scheduling by Chance
Constraint Programming
Bulent Erenay, PhD Candidate, Wilkes University,
5667 Barney Lane, Columbus, OH, 43235,
United States of America,
A stochastic patient scheduling problem is studied by using chance constraint
programming. The time patient stays at the hospital is considered as probablistic.
5 - Optimizing Ship Loading Schedules for Oil and Gas Terminals
Majid Algwaiz, Engineering Specialist, Saudi Aramco Oil
Company, P.O.Box 19422, Dhahran, 31311, Saudi Arabia,
Abdulaziz Nutaifi
We consider in this paper an oil and gas firm that owns its entire hydrocarbon
supply chain with many production facilities and ship loading terminals.
Customers make purchases a month in advance but only provide a four day
notice on the specific pickup times and the requested products and quantities. We
present a MILP formulation to manage the hydrocarbon network and assign ships
to berths on an hourly basis. Our objective is to minimize the ship waiting times
along with the demurrage fees.
WE03
03-Room 303, Marriott
Inventory Management - Inventory Policies
Contributed Session
Chair: Jim Shi, Assistant Professor, New Jersey Institute of Technology,
University Heights, Newark, NJ, 07102, United States of America,
1 - Stochastic Integrated Location-inventory Up-to-S Model in
Distribution System
Maxim Bushuev, Assistant Professor, Kent State University -
Geauga, 1835 Beacon Hill Cir #21, Cuyahoga Falls, OH, 44221,
United States of America,
A stochastic integrated location-inventory problem with up-to-S policy is
discussed. Simple proportional allocation rule is proposed which allows defining
and solving the problem as convex optimization. This is the first stochastic model
in the area of integrated location-inventory problems.
2 - An Extension of the Stochastic Dynamic Lot-Size Model of
Vargas to a Model with Uncertain Production
Hendrik Vermuyten, PhD Student, KU Leuven, Warmoesberg 26,
Brussel, 1000, Belgium,
We derive the optimal solution for the production planning for a single product
for every period in the planning horizon, when demand is stochastic and non-
stationary and the achievable production per period is stochastic as well. The
model is an adaption of the stochastic dynamic lot-size model of Vargas without
production restrictions. Simulation studies show a significant improvement in
expected costs for this model compared to the model of Vargas in case of
uncertain production capacity.
3 - Extended MIP Formulations for the Stochastic Lot-sizing Problem
Huseyin Tunc, Hacettepe University, Institute of Population
Studies, Sihhiye, Ankara, Turkey,
We revisit the certainty equivalent mixed integer programming formulations of
the stochastic lot-sizing problem under the static-dynamic uncertainty strategy,
and develop extended formulations thereof. The extended formulations are far
more time-efficient than the existing formulations in the literature. Also, instead
of working with a pre-determined piece-wise linear approximation of the cost
function, they can find a minimum cost solution by means of a novel dynamic cut
generation procedure.
4 - Stockout Risk Control of a Continuous
Production-inventory System
Jim Shi, Assistant Professor, New Jersey Institute of Technology,
University Heights, Newark, NJ, 07102, United States of America,
This paper studies the stockout control problem pertaining to a single-product
continuous-time production-inventory system with a constant replenishment
rate. Our objective is to optimize the expected system cost subject to a
predetermined stockout acceptance level.
WE03