WSC 2005

WSC 2005 Final Abstracts

PhD. Colloquium Track

Sunday 3:00:00 PM 6:00:00 PM
PhD Colloquium

Chair: Theresa Roeder (San Fransico State University)

Efficient Generation of Cycle Time-Throughput Curves through Simulation and Metamodeling
Feng Yang, Bruce E. Ankenman, and Barry L. Nelson (Northwestern University)

A cycle time-throughput (CT-TH) curve plays an important role in strategic planning for manufacturing systems. In this research, we seek to quantify the relationship of the moments and percentiles of cycle time to throughput rate via simulation experiments. The estimation of CT-TH moment curves is based on a nonlinear regression metamodel supported by queueing theory, and the adequacy of the model has been proved through our numerical experiments. Utilizing the estimated moment curves of cycle time, we proposed to estimate the CT-TH percentile curves in an indirect method assuming that the underlying distribution of cycle time is a generalized gamma, a highly flexible distribution. More specifically, we fit metamodels for the first three CT-TH moment curves throughout the throughput range of interest, determine the parameters of the generalized gamma by matching moments, and then obtain percentiles by inverting the distribution. To insure efficiency and control estimation error, simulation experiments are sequentially built up to optimize some design criterion.

A Real-Time Knowledge-Based Decission Support System for Health Care Quality Improvement Using Discrete Event Simulation Techniques
Alexander Komashie, Ali Mousavi, and Mustafa Ozbayrak (Brunel University )

The quality of health care is increasingly receiving more and more emphases in almost all developed countries. Traditional health care Quality Assurance (QA) models have always been implemented in retrospect depending heavily on surveys. The objective of this research is to develop a novel and more reliable approach to the monitoring and improvement of health care quality. This approach is based on a real-time simulation system that monitors a Health Care Quality Index (HCQI). This system will map the HCQI which is a function of process factors (e.g. waiting time), against the patients’ expectations and hence give health care managers the ‘local’ information needed to continuously adjust performance without waiting for an annual survey. Being a real-time system, managers will have the ability to run fast-forward simulation to help predict future demands and make decisions accordingly. This will be a key to Continuous Quality Improvement (CQI) in health care.

Scenario Planning for Simulation-Based Probabilistic Risk Assessment
Hamed Nejad (University of Maryland)

Simulation is used for the probabilistic risk assessment of complex systems that include hardware, software, and human elements. Since assessing the risk of such systems requires that a large number of scenarios be considered, a Planner component has been added to the simulation environment. This component solicits high level information such as system’s structure and functional behavior, and uses it to automatically generate and prioritize scenarios that will be used in risk assessment. Because of the hierarchical configuration of the Planner’s knowledge-base, scenarios can easily be modified to assess system risks when parts of the system are modified for risk management. As such, the analyst is able to compare the results of risk assessment -end-state probabilities as well as worse case scenarios- in different settings. The planning process is dynamic and simulation feedback is used to update the list of scenarios and/or their level of priority as needed.

Optimal Vehicle Scheduling & Layout for Automated Material Handling Systems(AMHS)
Sangwon Chae (University of California, Irvine)

This paper describes a scheduling of Automatic Material Handling System(AMHS) for improving productivity and reducing delivery time. In the paper, we simulate various scenarios and then analyze them for finding the optimal solution. In order to find the optimal solution, we use scenarios of vehicle scheduling and various layouts for AMHS in intra-bay. The paper shows the simulation to improve overall performance on AMHS by reducing the delivery time and raising productivity.

A Novel Approach to Studying Cell Communications Signalling
Jasmina Panovska (Heriot-Watt University)

In this paper we study the communication signals within a population of cells by extending the theory developed by Kummer and Ocone for the cell cycle. We show how the basic framework can be extrapolated from the cell level to the next organizational one, namely the population of cells. We define the basic quantities, such as the equivalent of the metabolic temperature and metabolic energy and we show how these are used to characterise the process of interest. We develop the basis for a non-equilibrium statistical theory for the system of cells and we establish the relationship between the variation in the local variables (eg. temperature and energy) and the variation in the cell concentration. The results from the theory are compared to those using the more classical approach of reaction-diffusion equations.

Estimation and Model Selection of the Interest Rates
Pouyan Mashayekh Ahangarani (USC)

A variety of continuous time series models of the short term riskless rate are estimated using Maximum Likelihood method on discretized models. Then the best model will be found that can fit the data better. A number of well-known models perform poorly in the comparison. Indirect Inference method is used for the best model in order to obtain consistent estimates. At the end, an empirical application of the stochastic model for interest rates will be used for pricing the call options of Nokia Company.

Quasi-Monte Carlo Simulation in a LIBOR Market Model
Xuefeng Jiang (Northwestern University) and John R. Birge (University Of Chicago)

Quasi-Monte Carlo (QMC) methods have been extensively applied to pricing complex derivatives. We apply quasi-Monte Carlo simulation to LIBOR market models based on the Brace-Gatarek-Musiela/Jamshidian framework. We price exotic interest-rate derivatives and compare the results using pseudo random sequences and different low discrepancy sequences.

Convergence of Strikes in Variance and Volatility Swaps
Ashish Jain (Columbia Business School)

In this work we study the convergence of strikes in variance and volatility swaps with frequency of sampling and determine the convergence rate of discrete time strikes to continuous time strikes. In our work, we study three different models of underlying evolution Black Scholes, Heston Stochastic Volatility model and Affine Jump diffusion processes to calculate the strike of the swap. First we determine the analytical value of the strike of variance swap in all these three models and then using convexity adjustment formula we calculate the approximate fair strike of volatility swap. For studying the convergence rate of strikes we compute the strikes for different sampling frequency using Monte Carlo Simulation. We found that convergence rate is linear which is also supported from the theoretical results we have proved.

What Value is Microsoft’s .Net to Discrete Event Simulations?
Adelaide Carvalho and Michael Pidd (Lancaster University)

Developers of simulation software have responded to the increasing demand for customised solutions by adding new features and tools to their simulation packages. This has led to monolithic applications with functionalities constantly extended by addition of templates, add-ons, etc in a generalising-customising-generalising development cycle. Though successful so far, this approach may be approaching its limit. An alternative approach is to compose simulation packages from prefabricated components that users may select, modify and assemble to acquire functionality to suit each model. This approach requires component-based paradigms and integration mechanisms. We investigate the value of .Net integration philosophy for development of discrete event simulation models. The DotNetSim prototype is used to investigate how software components developed within different packages can be linked into a single simulation application deployed as web services. DotNetSim consists of a graphical modelling environment and a base simulation engine.

Examining the Actual Benefits of Distributed Simulation for Supply Chain Management
Alexey Artamonov (Lancaster University)

Supply chains are not new, but their importance has grown in recent years due to globalisation, tough competition and the increasingly networked nature of business. Simulation has long been applied in production-inventory systems and more recently in supply chain management. Given that supply chains are distributed systems it seems sensible to consider the application of distributed computation to their simulation, as is evident from the number of recent research papers dedicated to distributed supply chain simulation (DSCS). However, attention is seldom paid to the actual advantages of this novel technology and the possible obstacles that must be overcome before real-world applications become commonplace. My research considers the potential drivers which might make this technology attractive and uses a DSCS testbed, implemented in AutoMod, to analyse whether these drivers can be achieved.

Adaptive Control Variates for American Option Pricing
Sam Ehrlichman and Shane G. Henderson (Cornell University)

Recently, Kim and Henderson (2004) have proposed a scheme for variance reduction in Monte Carlo simulations that uses stochastic optimization to compute an effective control variate. We apply this technique to the problem of American option pricing. While our work is similar in spirit to work by Bolia and Juneja (2005), the nonlinear procedure allows us to consider a more flexible class of approximating functions for our control variate; in particular, we are freed from having to make a particular choice of basis function parameterization up front.

Using Computer Simulations in Disaster Preparedness Exercises: A Study of Factors that Influence Learning and Performance in a First Responder Agency.
Daniel Joseph O'Reilly (Wayne State University)

This a dissertation study investigating the effectiveness of computer simulation as an instructional tool in disaster and mass emergency (including homeland security) preparedness with regard to learning and field response performance. The effectiveness will be determined through a summative performance evaluation of a local public health agency to a full-scale mock drill after receiving preparedness training through computer simulated instructional program(s). Simulation variables of fidelity, richness, prior experience of participants with emergencies, and prior experience with computer-simulation training will be included in the assessment. Currently, this study has evaluated applicable learning theories and the degree to which computer simulation meets the instructional criteria requirements proposed by these theories. A literature review of computer simulations in this instructional context has been found to be subjectively supportive but empirically weak. The objective of this study is to generate empirical evidence related to computer simulation in this current and particularly crucial arena of mass emergency preparedness and to contribute to the body of data on research in computer simulation instruction overall.

Multi-Hypothesis Intention Simulation Based Agent Architecture
Per M. Gustavsson (University of Skövde)

In the ongoing work with establishing Multi-Hypothesis Intention Simulation Based Agent Architecture the current status is presented. The intended architecture is described along with the Fractal Information Fusion Model. The work with establishing a framework for mapping environmental entities such as fire, wind into the Command and Control Information Exchange Data Model and the Coalition battle Management Language will be demonstrated. The current design and implementation of the mechanisms are discussed.

An Approach to Near Real-Time Dynamic Distributed System Control under Uncertainty
Kevin Adams (Virginia Tech)

Sensitivity analysis for integer optimization models is a very time-consuming process not conducive to applications requiring near real-time performance. The premise of this research is that the sensitivity of the optimization function in mixed integer non-linear systems creates a pattern. We demonstrate that a supervised neural network can be used to identify, classify and learn these patterns. The learned patterns can be represented in the various weights used in the neural network configuration. Using historical optimal information represented in the weights, the neural network can map the input constraints of our system to the optimal solution through functionally approximation with a high degree of accuracy. This functional approximation can be calculated in a small fixed determinate amount of time. In order to maintain good performance in a dynamic environment, the system uses feedback from an off-line component to identify and classify the constraint pattern changes. The proposed approach is demonstrated through simulation and case study to perform extremely well.

Interchanging Discrete Event Simulation Process-Interaction Models using the Web Ontology Language - OWL
Lee W. Lacy (University of Central Florida / DRC)

Discrete event simulation development requires significant investments in time and resources. Descriptions of discrete event simulation models are associated with world views, including the process-interaction orientation. Historically, these models have been encoded using high-level programming languages or special purpose (typically vendor-specific) simulation languages. These approaches complicate simulation model reuse and interchange. The current document-centric World Wide Web is evolving into a Semantic Web that communicates information using ontologies. The Web Ontology Language – OWL, was used to encode an ontology for representing discrete event process-interaction models (DEPIM). The DEPIM ontology was developed using ontology engineering processes. The purpose of DEPIM is to provide a vendor-neutral open representation to support model interchange. Model interchange provides an opportunity to improve simulation quality, reduce development costs, and reduce development times.

Analysis of Production Authorization Card Schemes using Simulation and Neural Network Metamodels
Corinne MacDonald and Eldon A. Gunn (Dalhousie University)

We have developed a framework to model and analyze the performance of complex manufacturing systems operating under a variety of production control strategies. This framework involves a production authorization card scheme, which enables emulation of many popular strategies such as kanban or Base Stock systems. A discrete-event simulation model of the manufacturing system produces estimates of the multiple system performance measures, such as average work-in-process inventory and customer service rates, for combinations of control parameters. Finally, neural network metamodels are trained to approximate the expected value of these system performance measures, using a subset of parameter combinations and the corresponding performance estimates generated by the simulation model. We will show that this framework provides a flexible means of conducting analysis of the impact of parameter settings on the performance of the system, and is a viable alternative to simulation optimization.

The Use of Hyper-Hidden Markov Models in Wireless Channel Simulation
Antonia Marie Boadi (University of Southern California)

The transmission of multimedia over satellite, cellular or mobile wireless networks introduces impairments which degrade signal quality. This thesis proposes a channel simulation model and design philosophy that represents a shift from the use of traditional network-centric design requirements to a more comprehensive approach that encompasses perceptual quality issues. PerFEC, the Perceptually-Sensitive Forward Error Control agent, employs a flexible, adaptive coding scheme that is both quality-of-service (QoS) and quality-of-perception (QoP) configurable. PerFEC is a decision agent that uses decoder output to select an error control scheme that is appropriate for the prevailing channel conditions. The channel simulation model is based on a new construct, the Hyper-Hidden Markov Model (HHMM), whose internal states have dual interpretation: they represent the user's perceptual quality as well as an estimate of the current bit-error-rate. This layering of Hidden Markov Models provides insight into the nonlinear relationship between quality-of-service metrics and perceptual quality.

Simulation Modeling of the Level of Use of E-Health System and Optimization of its Effect on Patient Quality of Life
Abhik Bhattacharya and David H. Gustafson (University of Wisconsin-Madison)

Comprehensive Health Enhancement Support System (CHESS) is a disease-specific computer-based system designed to meet information and support needs. The users of discussion group, the most used service provided through CHESS, gain enhanced quality of life (QoL). A benefit-based model for sustainable use of discussion group is developed, validated, and analyzed. The model deals with the system of opposing forces that link discussion group size and communication activity and the chances of participation by a member. The model was calibrated based on a subset of empirical data collected from two randomized clinical trials of breast cancer patients, while the remaining data was used to validate. Simulation experiments were conducted to determine the model’s predictions regarding the impact of CHESS on QoL. The results imply that the use of communication in electronic support groups will enhance the QoL of cancer patients balancing the opposing forces of group size and communication activity.

Selecting the Best System and Determining a Set of Feasible Systems
Demet Batur (Georgia Institute of Technology, Industrial and Systems Engineering)

We present two fully sequential indifference-zone procedures to select the best system from a number of competing simulated systems, where best is defined by maximum or minimum expected performance. These two procedures have parabolic continuation regions rather than triangular continuation regions employed in a number of papers. Our procedures find the best or near-best system with at least a pre-specified probability of correctness when basic observations are independent and identically normally distributed. They allow for unequal and unknown variances across systems and the use of common random numbers, and show moderate improvement compared to other existing fully sequential procedures with triangular continuation regions. We also present procedures for finding a set of feasible or near-feasible systems with some statistical guarantee among a finite number of simulated systems in the presence of stochastic constraints on secondary performance measures. The proposed procedures can handle a large number of systems and stochastic constraints.