K.P. White

A comparison of five steady-state truncation heuristics for simulation

We compare the performance of five well-known truncation heuristics for mitigating the effects of initialization bias in the output analysis of steady-state simulations. Two of these rules are variants of the MSER heuristic studied by White (1997); the remaining rules are adaptations of bias-detection tests based on the seminal work of Schruben (1982). Each heuristic was tested in each of 168 different experiments. Each experiment comprised multiple tests on different realizations of the sample path of a second-order autoregressive process with known (deterministic) bias function. Different experiments employed alternative process parameters, generating a range of damped and underdamped stochastic responses. These were combined with alternative damped, underdamped, and mean shift bias functions. The performance of each rule was evaluated based on the ability of the rule to remove bias from the mean estimator for the steady-state process. Results confirmed that four of the five rules were effective and reliable, consistently yielding truncated sequences with reduced bias. In general, the MSER heuristics outperformed the three rules based on bias detection, with Spratts (1998) MSER-5 the most effective and robust choice for a general-purpose method.

A survey of data resources for simulating patient flows in healthcare delivery systems

Modeling and simulation studies of patient flows in healthcare systems have been reported consistently in these proceedings for over a decade. Our ongoing research in this area is motivated by our desire to develop models which will illuminate the causes and remedies for repeated area-wide ambulance diversions experienced in a metropolitan hospital system. In this paper we summarize our background research on the sources of data available to calibrate patient-flow simulation models, including time series for patient admission, discharge, diagnoses, length-of-stay, and inpatient census for emergency departments and hospitals. Specifically, we review the input analyses reported for various prior simulation studies, including data capture and technical difficulties in reducing data for model calibration. We also suggest alternative sources of data that could prove especially useful in simulation studies of mass ambulance diversions, as well as heavy, area-wide patient loads that might be associated with emergency responses.

Using RFID technologies to capture simulation data in a hospital emergency department

Simulation professionals understand the importance of accurate data for model validation. Traditional sources of simulation data come from information technology systems, manual records from staff, observations, and estimates by subject matter experts. This paper discusses how radio frequency identification (RFID) technologies were used on a recent consulting engagement at a hospital. Data collected through RFID can validate or replace activity duration estimates from traditional sources. However, the accuracy and cost effectiveness of RFID is not guaranteed. A sound methodology was developed, which included rigorous planning and testing of hardware, processes and data analysis. Hardware vendors needed to understand what the simulation required so they could properly setup equipment and software. Also, ED staff needed to understand the purpose of this data collection to avoid anxiety about personnel evaluations. Finally, efficient and reliable issue and collection of patient tags was crucial to the success of this effort

Stationarity tests and MSER-5: exploring the intuition behind mean-squared-error-reduction in detecting and correcting initialization bias

We explore the reasoning behind MSER-5, an efficient and effective truncation heuristic for reducing initialization bias in steady-state simulation. We also compare MSER-5 with the KPSS stationarity test as one means of investigating the possibility that MSER¿s effectiveness is the result of its utility as a stationarity measure. Conversely, this comparison also lets us explore whether or not a stationarity test from the time-series literature can be used as an effective initialization bias-control heuristic. Finally, we investigate the use of an alternative form of MSER-5 that uses a variance estimator that adjusts for serial correlation.

Stochastic approximation with simulated annealing as an approach to global discrete-event simulation optimization

This paper explores an approach to global, stochastic, simulation optimization which combines stochastic approximation (SA) with simulated annealing (SAN). SA directs a search of the response surface efficiently, using a conservative number of simulation replications to approximate the local gradient of a probabilistic loss function. SAN adds a random component to the SA search, needed to escape local optima and forestall premature termination. Using a limited set of simple test problems, we compare the performance of SA/SAN with the commercial package OptQuest. Results demonstrate that SA/SAN can outperform OptQuest when properly tuned. The practical difficulty lies in specifying an appropriate set of SA/SAN gain coefficients for a given application. Further results demonstrate that a multistart approach greatly improves the coverage and robustness of SA/SAN, while also providing insights useful in directing iterative improvement of the gain coefficients before each new start. This preliminary study is sufficiently encouraging to invite further research on SA/SAN.

Determining a warm-up period for a telephone network routing simulation

We present a new approach to determining the warm-up period for steady-state simulation of telephone traffic. The underlying simulation model captures the sophisticated interactions that determine the acceptance and routing of calls between origin and destination nodes across the telephone network. Recognizing that both the arrival and duration of calls are Markovian, approximate satisfaction of the equivalence property of Jackson networks signifies a stochastic steady state. We are able to determine the onset of steady-state behavior, therefore, by monitoring arrival and departure rates observed during the simulation and testing for equivalence. Application of the rule is illustrated using a simple three-node network.

Operational simulation of an X-ray lithography cell: comparison of 200 mm and 300 mm wafers

We review progress on a project to evaluate prospective operations in a semiconductor wafer fab that employs next generation, proximity X-ray lithography to pattern the critical dimensions of computer chips. A simulation model is developed that captures the processing of wafers through an X-ray lithography cell using a synchrotron as the source of exposure radiation. The model incorporates the best current information on unit-cell design and processing times and implements a range of events that interrupt the flow of wafers processing on the cell. Performance measures estimated from the simulation include the weekly throughput for the cell and the frequency of SEMI E-10 equipment states for the corresponding exposure tool. Simulation experiments are conducted to compare the performance of a cell fabricating 200 mm wafers with that of a cell fabricating 300 mm wafers, for each of three different chip sizes. Results illustrate the anticipated dependence of average wafer throughput on wafer size and assumptions regarding the number of chips per wafer, with a maximum of approximately 3400 wafers/week for 200 mm wafers with 25/spl times/25 mm field size. Ignoring wafer-sort losses, however, a maximum throughput of approximately 410,000 chips/week is realized for 300 mm wafers with 11/spl times/22 mm fields. Remarkably, the distribution of equipment states remains relatively unchanged across simulation experiments.

A Bayesian approach to analysis of limit standards

Limit standards are probabilistic requirements or benchmarks regarding the proportion of replications conforming or not conforming to a desired threshold. Sample proportions resulting from the analysis of replications are known to be beta distributed. As a result, standard constructs for defining a confidence interval on such a proportion, based on critical points from the normal or Students t distribution, are increasingly inaccurate as the mean sample proportion approaches the limits of 0 or 1. We consider the Bayesian relationship between the beta and binomial distributions as the foundation for a sequential methodology in the analysis of limit standards. The benefits of using the beta distribution methodology are variance reduction, and smaller sample size (when compared to other analysis methodologies).

Comparison of Bayesian priors for highly reliable limit models

Limit standards are probability interval requirements for proportions. Simulation literature has focused on finding the confidence interval of the population proportion, which is inappropriate for limit standards. Further, some frequentist approaches cannot be utilized for highly reliable models, or models which produce no or few non-conforming trials. Bayesian methods provide approaches that can be utilized for all limit standard models. We consider a methodology developed for Bayesian reliability analysis, where historical data is used to define the a priori distribution of proportions p, and the customer desired a posteriori maximum probability is utilized to determine sample size for a replication.

Architectural concepts for a system simulator for concurrent prototyping of equipment and controls

AutoMod/sup /spl reg// is a leading discrete-event simulation package widely applied in the modeling and analysis of distribution systems. Included in the AutoMod software suite is the Model Communications Module (MCM), which allows an executing simulation to open socket connections and to send and receive messages via TCP/IP network protocol. In this paper we report on a pilot study which explores the functionality of the MCM. In particular, we develop and implement an architecture that can be used to design, test, verify, and optimize control system software interacting with a discrete-event simulation of the system to be controlled. This architecture supports concurrent engineering of controls and hardware prototypes. Application of this architecture can significantly reduce the duration and cost of development cycles for new equipment and systems. In addition, this architecture can be applied to investigate the feasibility of implementing engineering changes in systems currently deployed.