DISCRETE EVENT SIMULATION MODELING FOR RAMI ANALYSIS OF NORMAL AND UPSET PLANT OPERATION
J. J. Nutaro and J. C. Schryver Computational Sciences and Engineering Division Oak Ridge National Laboratory* P.O. Box 2008 Oak Ridge, Tennessee, 37831-6058 USA M. J. Haire Fuel Cycle and Isotopes Division Oak Ridge National Laboratory* P.O. Box 2008 Oak Ridge, Tennessee 37831-6179 USA
For the American Nuclear Society Summer Meeting Hollywood, Florida June 26–30, 2011
The submitted manuscript has been authored by a contractor of the U.S. Government under contract DE-AC05-00OR22725. Accordingly, the U.S. Government retains a nonexclusive, royalty-free license to publish or reproduce the published form of this contribution, or allow others to do so, for U.S. Government purposes.
_________________________ *
Oak Ridge National Laboratory, managed by UT-Battelle, LLC, for the U.S. Department of Energy under contract DE-AC05-00OR22725.
Discrete Event Simulation Modeling for RAMI Analysis of Normal and Upset Plant Operation*
J. J. Nutaro and J. C. Schryver Computational Sciences and Engineering Division, Oak Ridge National Laboratory P.O. Box 2008, Oak Ridge Tennessee 37831-6058, e-mail:
[email protected] and
[email protected] M. J. Haire Fuel Cycle and Isotopes Division, Oak Ridge National Laboratory P.O. Box 2008, Oak Ridge, Tennessee 37831-6179, e-mail:
[email protected]
INTRODUCTION A nuclear fuel cycle facility has dozens of systems, hundreds of subsystems and components, and tens of thousands of parts. There are numerous closely related ways to model the performance of such a facility. Popular methods include event trees, fault trees, reliability block diagrams, and Markov state diagrams. Though such techniques are useful in many circumstances, it is difficult with these modeling approaches to describe time dependence in a natural way. However, the operation of high-availability systems can depend critically on how failure and recovery unfold with time. To address the timedependent aspect of these problems, discrete event simulation is an effective, flexible, and efficient approach to modeling and analysis. This paper summarizes use of discrete event simulation to assess the availability and throughput of a nuclear fuel cycle facility in the wake of upset conditions. SIMULATION ANALYSIS Simulation models are structured in the time domain, where the flow of events can be observed and evaluated. In discrete event simulations, time advances in intervals that span the start and completion of events (see, e.g., Ref. [1]). When a discrete event model is built for reliability, availability, maintainability, and inspectability (RAMI) analysis, tasks and operations are modeled as discrete, chronologically ordered steps. This set of steps may include decision points at which the workflow branches. For example, the decision to use machine A or machine B may depend on the status of the machines when they are called for in the workflow. The result of this modeling process is a block diagram that captures key aspects of the facility’s operation: for
example, it identifies resources shared by tasks and alternative workflows and relates the dependence of tasks on resources through time. The discrete event model is programmed with the support of an appropriate simulation package. Simulations runs are made to assess the impact of time-dependent equipment operations—failure, repairs, and patterns of usage—on the performance of the system. Typically, the simulator’s input includes RAMI information (e.g., time to failure, repair times), the duration of work steps, and other elements that affect equipment wear and reliability. The simulation moves entities through their workflows until a failure occurs in some system component. After the failure is repaired, the simulation continues until the next failure. Performance measures (e.g., number of failures and total downtime) are collected during the simulation run. Because the model uses random variables to describe time to failure, time to repair, and time to complete an operation, it is essential to obtain a statistically significant number of simulation runs before drawing conclusions (see, e.g., Ref. [2]). In constructing a simulation model, it is therefore necessary to balance a desire for detailed descriptions of the system with the requirement to perform hundreds of simulation runs to draw statistically meaningful conclusions about the impacts of alternative designs. AN EXAMPLE: THE IMPACT ON THROUGHPUT OF STANDBY MACHINES A technique for improving plant availability and throughput is to have redundancy for failure-prone equipment. Consider the redundant machines as shown in Fig. 1a. Incoming work is queued at the switch, which feeds jobs to the primary machine or, if the primary machine is inoperable, to the spare machine. The dotted lines show partially completed or off-specification product that is
*Notice: This manuscript has been authored by UT-Battelle, LLC, under contract DE-AC05-00OR22725 with the U.S. Department of Energy. The United States Government retains and the publisher, by accepting the article for publication, acknowledges that the United States Government retains a non-exclusive, paid-up, irrevocable, world-wide license to publish or reproduce the published form of this manuscript, or allow others to do so, for United States Government purposes.
returned to the queue for rework. The endogenous availability of this system is the ratio of its observed production rate to its ideal production rate. Figure 1b shows the impact of using an inferior spare machine, which is modeled by a spare with a production rate that is a fraction of the primary machine’s rate. Simulation reveals that endogenous availability is linearly related to the quality of the spare. This relationship can be discovered only by a dynamic analysis of the system.
Primary Switch Spare
SUMMARY AND CONCLUSIONS We have spent the past 4 years applying discrete event simulation to estimate the availability and production capacity of a nuclear material production facility. In the course of this work, we have uncovered several issues that are discoverable only through a dynamic analysis. For example, it was found that specific temporal relationships between failures in cyclically connected processes can induce gridlock in the production system. These types of gridlock conditions had been observed in similar older facilities but were not thought to be a potential problem in the new design. This discovery early in the design process prevented expensive retrofits later in the plant’s life cycle. In this and several other instances, dynamic analysis of plant operations has had a substantial impact on design. Such impacts have repeatedly justified investment in this analysis technique. REFERENCES 1. B. P. ZIEGLER, H. PRAEHOFER, and T. G. KIM, Theory of Modeling and Simulation, 2nd ed., Academic Press, San Diego (2000). 2. A. M. LAW and W. D. KELTON, Simulation Modeling and Analysis, 2nd ed., McGraw-Hill, New York (1991).
Fig. 1: (a) Primary and spare machine in parallel. (b) Endogenous availability as a function of the spare’s production rate.
