|HOME||Past Research Interests
Please note: I am retired from McMaster University and cannot take on any new graduate students at this time. This page is retained for archival purposes.
To support the McMaster Nuclear Reactor (MNR), the following needs have been identified:
Possible new research topics focused on MNR include:
Safety analysis on nuclear reactors shows that 70% of all nuclear accidents have a root cause related to Human Factors (HF), the remainder being due mainly to mechanical failures. Of the mechanical failures, the large accident contribution is small compared to minor system failures. Both HF failures and mechanical failures are strongly influenced by design and analysis. In turn, design and analysis is strongly influenced by the design and analysis process (i.e. tools and methodology) which is largely a hands on activity for the design engineer and the technical support unit. Empowerment of the designer, the technical support unit and the operator (resulting from providing the right tools, from providing job satisfaction, from providing a proper safety culture, etc.) is arguably the single most important factor influencing nuclear reactor risk in particular and hence is a key factor in the design of PSS for engineering. The predominant design and analysis tool in reactor physics and thermalhydraulics is the large monolithic computer code (e.g. WIMS or CATHENA) which is invariably treated as a black box by all except the code designer and a few select analysts. This leaves the designer / analyst at "arms length" from the engineering task. These large codes are used for both scoping studies and for final analyses. The large codes remain essential for the final analyses but it is hypothesized that design and analysis scoping studies will be considerably enhanced by the use of simple, targeted small codes that are constructed by the designer from an industry-wide library of routines that have undergone quality assurance and quality control testing. This will allow, in fact requires, the user to be tightly integrated into the engineering activities, making it "hands on". The first general objective of the proposed research is thus directed towards Do It Yourself or DIY agents in reactor physics and thermalhydraulics. The construction and testing of the individual agents are straightforward. In-house code exists for water properties and for thermalhydraulic simulation. The correlation library needs to be extended extensively. An in-house core physics code exists but it needs to be extended in its property library and simulation of transients. One Ph.D. student, Dave Gilbert, has worked on a Reactor Simulation Editor geared to provide a robust environment for DIY agents in reactor physics core modelling.
The existing standard industrial codes are qualified for use on large power reactors. Increasingly, these codes are being called upon for use on small cores such as research and isotope production reactors. Herein, the research objectives relating to reactor physics are to prove, with respect to research reactors such as the McMaster Nuclear Reactor that diffusion theory is adequate for core analysis, fuel substitution predictions and control rod calibration predictions, that the computer codes Dragon and WIMS give accurate cell parameters, that a recipe for the use of these codes on research reactors can be developed, and that analytical error estimates of diffusion theory can be provided. To achieve these objectives, Dragon and WIMS will be compared to in-core measurements, to MCNP and to accepted bench mark problems, and perturbation theory will be pursued as a means of providing an analytical estimation of diffusion theory errors. In the area of thermalhydraulics, the objective is to prove the adequacy of the codes and of the current MNR design.
There exists a substantial experimental database for MTR type reactors, the SPERT and BORAX tests. The McMaster Nuclear Reactor is a MTR type reactor. The research objective is to glean as much information from these tests to support the MNR Safety Analysis Report. One Ph. D. student, now graduated, has completed working on this project.
A perennial reactor physics question is whether a full 3-D reactor physics model is needed for fast transient analysis of the MNR core or whether a point model will suffice.
To support reactor operations and safety analysis, MNR needs to be characterized both thermalhydraulically and neutronically.
The detailed velocity distributions in the MNR fuel assemblies have been measured
experimentally using laser-doppler anemometry by T.S. Ha at the Master's level.
The next steps are to:
- characterize the core bypass flow,
- investigate the influence of the entrance effects on assembly flow,
- determine air cooling limits of the fuel.
Probabilistic Safety Analysis (PSA) has been conducted for the McMaster Nuclear Reactor. This PSA work and deterministic analysis form the basis for the current Safety Analysis Report, required for licensing. To support this work, event trees and fault trees are constructed for 4 accident events suspected of being the main contributors to risk (LOCA, LOR, loss of flow and flow blockage). Current PSAs have a number of limitations that need to be addressed, as indicated below.
An estimation of nonevent probabilities can be made using Bayesian statistics and other statistical techniques. A sensitivity analysis needs to be conducted to determine the acceptability of the Bayesian approach.
Conventional PSA methodology indicate that dynamic effects with long time constants can be addressed within the existing framework. Time-dependent effects with long time constants can be accommodated by modifying the estimates of the probabilities of the basic events in the ET/FT models, development of logic models for different plant configurations, and periodic updating of the PSA models (living PSA). However, only a limited, implicit treatment of the dynamic effects with short time constants is possible with conventional PSA methodologies because the conventional PSAs are limited in their ability to represent the interactive nature of the roles of the systems. One Ph.D. student has investigated the sensitivity of risk estimates to the short term dynamic effect.
The log term goal is to build a Risk Meter by adding failure models to the MNR Simulator (below). The intent is to use the simulator to build up an event history that, in effect, defines the risk profile without the need to hand-create event trees and fault trees.
Reactor Physics Space-time Kinetics
The multigrid method, normally applied to thermalhydraulic simulation, has been found to be applicable to neutron space-time kinetics. One Ph.D. student has investigated.
Joint with Dr. W.F. Skip Poehlman. The hardware part includes the build of an intelligent data acquisition system able to extract the Iodine signal from the reactor background. The software part includes the distributed application software. The software is written in Delphi4 and the issue is to create a graphic user interface able to adapt to user requirements. The improved system has been designed and implemented. Several years of operation have proven the design but there are opportunities for extension of the system. A Master's level student rebuilt the system using LabView.
Maintenance Optimization needs to be investigated. The current focus is on the control system for MNR.
Control Strategies for Loosely-Coupled Architectures: The objective of the research related to PSS control strategies is to prove that a suitable control strategy for a loosely coupled system of autonomous computing agents can be constructed by marrying known control algorithms to the agents based on statically and dynamically determined agent characteristics. The proposed approach is to determine a likely control model to be used with nonlinear multi-variable control by performing a comparison of the dominant control models in use currently. The next step would be to try to use such a controller on relatively simple computation problems and determine what the numerical equivalents are to control parameters on physical systems; e.g. error, setpoint, etc. Ph.D. student Carl Schwellnus is currently investigating the above.
MNRSIM: In the past, the individual research performed by students tended to be somewhat isolated and self-contained. Subsequent students tended to not benefit from the earlier works, especially in the area of computer codes. Consequently, a computing environment was sought that uses standard C, has a graphical library, supports a simulator interface, multitasking and distributed computing. LabWindows/CVI from National Instruments was chosen as the computing platform. MNRSIM is under construction. The intent is to use it as the common modelling platform for the bulk of the research in this group so that future students can build on past work and so that a full-scope simulator eventually emerges to support operator training and technical analysis support for MNR.