Upgrading automation for nuclear fuel in-core management: from the symbolic generation of configurations to the neural adaptation of heuristics

FUELCON is an expert system in nuclear engineering. Its task is optimized refueling-design, which is crucial to keep down operation costs at a plant. FUELCON proposes sets of alternative configurations of fuel-allocation; the fuel is positioned in a grid representing the core of a reactor. The practitioner of in-core fuel management uses FUELCON to generate a reasonably good configuration for the situation at hand. The domain expert, on the other hand, resorts to the system to test heuristics and discover new ones, for the task described above. Expert use involves a manual phase of revising the ruleset, based on performance during previous iterations in the same session. This paper is concerned with a new phase: the design of a neural component to carry out the revision automatically. Such an automated revision considers previous performance of the system and uses it for adaptation and learning better rules. The neural component is based on a particular schema for a symbolic to recurrent-analogue bridge, called NIPPL, and on the reinforcement learning of neural networks for the adaptation.

A novel neural/symbolic hybrid approach to heuristically optimized fuel allocation and automated revision of heuristics in nuclear engineering

FUELCON is an expert system for optimized refueling design in nuclear engineering. This task is crucial for keeping down operating costs at a plant without compromising safety. FUELCON proposes sets of alternative configurations of allocation of fuel assemblies that are each positioned in the planar grid of a horizontal section of a reactor core. Results are simulated, and an expert user can also use FUELCON to revise rulesets and improve on his or her heuristics. The successful completion of FUELCON led this research team into undertaking a panoply of sequel projects, of which we provide a meta-architectural comparative formal discussion. In this paper, we demonstrate a novel adaptive technique that learns the optimal allocation heuristic for the various cores. The algorithm is a hybrid of a fine-grained neural network and symbolic computation components. This hybrid architecture is sensitive enough to learn the particular characteristics of the ‘in-core fuel management problem’ at hand, and is powerful enough to use this information fully to automatically revise heuristics, thus improving upon those provided by a human expert.

The FUELCON meta-architecture, in the landscape of intelligent technologies for refuelling

The conception of the FUELCON architecture, of a composite tool for the generation and validation of patterns for assigning fuel assemblies to the positions in the grid of a reactor core section, has undergone an evolution throughout the history of the project. Different options for various subtask were possible, envisioned, or actually explored or adopted. We project these successive, or even concomitant configurations of the architecture, into a meta-architecture, which quite not by chance happens to reflect basic choices in the field's history over the last decade.

Refueling in nuclear engineering: the FUELCON project

Over a time span of almost a decade, the FUELCON project in nuclear engineering has led to a fully functional expert system and spawned sequel projects. Its task is in-core fuel management, also called `refueling', i.e., good fuel-allocation for reloading the core of a given nuclear reactor, for a given operation cycle. The task is crucial for keeping down operation costs at nuclear power plants. Fuel comes in different types and is positioned in a grid representing the core of a reactor. The tool is useful for practitioners but also helps the expert in the domain to test his or her rules of thumb and to discover new ones.

Three-dimensional free surface modelling in an unstructured mesh environment for metal processing applications

In the casting of metals, tundish flow, welding, converters, and other metal processing applications, the behaviour of the fluid surface is important. In aluminium alloys, for example, oxides formed on the surface may be drawn into the body of the melt where they act as faults in the solidified product affecting cast quality. For this reason, accurate description of wave behaviour, air entrapment, and other effects need to be modelled, in the presence of heat transfer and possibly phase change. The authors have developed a single-phase algorithm for modelling this problem. The Scalar Equation Algorithm (SEA) (see Refs. 1 and 2), enables the transport of the property discontinuity representing the free surface through a fixed grid. An extension of this method to unstructured mesh codes is presented here, together with validation. The new method employs a TVD flux limiter in conjunction with a ray-tracing algorithm, to ensure a sharp bound interface. Applications of the method are in the filling and emptying of mould cavities, with heat transfer and phase change.

Monotone scheme and boundary conditions for finite volume simulation of magnetohydrodynamic internal flows at high Hartmann number

A monotone scheme for finite volume simulation of magnetohydrodynamic internal flows at high Hartmann number is presented. The numerical stability is analysed with respect to the electromagnetic force. Standard central finite differences applied to finite volumes can only be numerically stable if the vector products involved in this force are computed with a scheme using a fully staggered grid. The electromagnetic quantities (electric currents and electric potential) must be shifted by half the grid size from the mechanical ones (velocity and pressure). An integral treatment of the boundary layers is used in conjunction with boundary conditions for electrically conducting walls. The simulations are performed with inhomogeneous electrical conductivities of the walls and reach high Hartmann numbers in three-dimensional simulations, even though a non-adaptive grid is used.

Converting best manual practice into generic automatable strategies for unstructured mesh parallelisation

The manual effort required to convert sequential computational mechanics programs into a useful, scalable parallel form is considerable. Tools that can assist in the conversion process are clearly required. Computer aided parallelisation tools (CAPTools) have been developed to generate efficient parallel code for real world structured grid application codes such as Computational Fluid Dynamics. Automatable single-program multi-data (SPMD) overlapping domain decomposition (DD) techniques established for structured grid codes have been adapted by the authors to manually parallelise unstructured mesh applications. Inspector loops have been used to provide generic techniques for the run-time support necessary to extend the capabilities of CAPTools to automatic implementation of SPMD DD techniques in the parallelisation of unstructured mesh codes. Copyright © 1999 John Wiley & Sons, Ltd.

The implementation of an aeronautical CFD flow code onto distributed memory parallel systems

The parallelization of an industrially important in-house computational fluid dynamics (CFD) code for calculating the airflow over complex aircraft configurations using the Euler or Navier–Stokes equations is presented. The code discussed is the flow solver module of the SAUNA CFD suite. This suite uses a novel grid system that may include block-structured hexahedral or pyramidal grids, unstructured tetrahedral grids or a hybrid combination of both. To assist in the rapid convergence to a solution, a number of convergence acceleration techniques are employed including implicit residual smoothing and a multigrid full approximation storage scheme (FAS). Key features of the parallelization approach are the use of domain decomposition and encapsulated message passing to enable the execution in parallel using a single programme multiple data (SPMD) paradigm. In the case where a hybrid grid is used, a unified grid partitioning scheme is employed to define the decomposition of the mesh. The parallel code has been tested using both structured and hybrid grids on a number of different distributed memory parallel systems and is now routinely used to perform industrial scale aeronautical simulations. Copyright © 2000 John Wiley & Sons, Ltd.

Three parallel programming paradigms: Comparisons on an archetypal PDE computation

Three paradigms for distributed-memory parallel computation that free the application programmer from the details of message passing are compared for an archetypal structured scientific computation -- a nonlinear, structured-grid partial differential equation boundary value problem -- using the same algorithm on the same hardware. All of the paradigms -- parallel languages represented by the Portland Group's HPF, (semi-)automated serial-to-parallel source-to-source translation represented by CAP-Tools from the University of Greenwich, and parallel libraries represented by Argonne's PETSc -- are found to be easy to use for this problem class, and all are reasonably effective in exploiting concurrency after a short learning curve. The level of involvement required by the application programmer under any paradigm includes specification of the data partitioning, corresponding to a geometrically simple decomposition of the domain of the PDE. Programming in SPMD style for the PETSc library requires writing only the routines that discretize the PDE and its Jacobian, managing subdomain-to-processor mappings (affine global-to-local index mappings), and interfacing to library solver routines. Programming for HPF requires a complete sequential implementation of the same algorithm as a starting point, introduction of concurrency through subdomain blocking (a task similar to the index mapping), and modest experimentation with rewriting loops to elucidate to the compiler the latent concurrency. Programming with CAPTools involves feeding the same sequential implementation to the CAPTools interactive parallelization system, and guiding the source-to-source code transformation by responding to various queries about quantities knowable only at runtime. Results representative of "the state of the practice" for a scaled sequence of structured grid problems are given on three of the most important contemporary high-performance platforms: the IBM SP, the SGI Origin 2000, and the CRAYY T3E.

Viscoelastic flow through a planar contraction using a semi-Lagrangian finite volume method

A semi-Lagrangian finite volume scheme for solving viscoelastic flow problems is presented. A staggered grid arrangement is used in which the dependent variables are located at different mesh points in the computational domain. The convection terms in the momentum and constitutive equations are treated using a semi-Lagrangian approach in which particles on a regular grid are traced backwards over a single time-step. The method is applied to the 4 : 1 planar contraction problem for an Oldroyd B fluid for both creeping and inertial flow conditions. The development of vortex behaviour with increasing values of We is analyzed.

SMARTFIRE: an integrated computational fluid dynamics code and expert system for fire field modelling

SMARTFIRE, an open architecture integrated CFD code and knowledge based system attempts to make fire field modeling accessible to non-experts in Computational Fluid Dynamics (CFD) such as fire fighters, architects and fire safety engineers. This is achieved by embedding expert knowledge into CFD software. This enables the 'black-art' associated with the CFD analysis such as selection of solvers, relaxation parameters, convergence criteria, time steps, grid and boundary condition specification to be guided by expert advice from the software. The user is however given the option of overriding these decisions, thus retaining ultimate control. SMARTFIRE also makes use of recent developments in CFD technology such as unstructured meshes and group solvers in order to make the CFD analysis more efficient. This paper describes the incorporation within SMARTFIRE of the expert fire modeling knowledge required for automatic problem setup and mesh generation as well as the concept and use of group solvers for automatic and manual dynamic control of the CFD code.

A knowledge-based system to represent spatial reasoning for fire modelling

This paper describes the architecture of the knowledge based system (KBS) component of Smartfire, a fire field modelling tool for use by members of the fire safety engineering community who are not expert in modelling techniques. The KBS captures the qualitative reasoning of an experienced modeller in the assessment of room geometries, so as to set up the important initial parameters of the problem. Fire modelling expertise is an example of geometric and spatial reasoning, which raises representational problems. The approach taken in this project is a qualitative representation of geometric room information based on Forbus’ concept of a metric diagram. This takes the form of a coarse grid, partitioning the domain in each of the three spatial dimensions. Inference over the representation is performed using a case-based reasoning (CBR) component. The CBR component stores example partitions with key set-up parameters; this paper concentrates on the key parameter of grid cell distribution.

An experimental and numerical CFD study of turbulence in a tundish container

This paper describes work performed at IRSID/USINOR in France and the University of Greenwich, UK, to investigate flow structures and turbulence in a water-model container, simulating aspects typical of metal tundish operation. Extensive mean and fluctuating velocity measurements were performed at IRSID using LDA to determine the flow field and these form the basis for a numerical model validation. This apparently simple problem poses several difficulties for the CFD modelling. The flow is driven by the strong impinging jet at the inlet. Accurate description of the jet is most important and requires a localized fine grid, but also a turbulence model that predicts the correct spreading rates of jet and impinging wall boundary layers. The velocities in the bulk of the tundish tend to be (indeed need to be) much smaller than those of the jet, leading to damping of turbulence, or even laminar flow. The authors have developed several low-Reynolds number (low-Re) k–var epsilon model variants to compute this flow and compare against measurements. Best agreement is obtained when turbulence damping is introduced to account not only for walls, but also for low-Re regions in the bulk – the k–var epsilon model otherwise allows turbulence to accumulate in the container due to the restricted outlet. Several damping functions are tested and the results reported here. The k–ω model, which is more suited to transitional flow, also seems to perform well in this problem.

Modelling the fluid dynamics and coupled phenomena in arc weld pools

A 3D model of melt pool created by a moving arc type heat sources has been developed. The model solves the equations of turbulent fluid flow, heat transfer and electromagnetic field to demonstrate the flow behaviour phase-change in the pool. The coupled effects of buoyancy, capillary (Marangoni) and electromagnetic (Lorentz) forces are included within an unstructured finite volume mesh environment. The movement of the welding arc along the workpiece is accomplished via a moving co-ordinator system. Additionally a method enabling movement of the weld pool surface by fluid convection is presented whereby the mesh in the liquid region is allowed to move through a free surface. The surface grid lines move to restore equilibrium at the end of each computational time step and interior grid points then adjust following the solution of a Laplace equation.

Predictions of segregation of granular material with the aid of PHYSICA, a 3-D unstructured finite-volume modelling framework

In this paper a continuum model for the prediction of segregation in granular material is presented. The numerical framework, a 3-D, unstructured grid, finite-volume code is described, and the micro-physical parametrizations, which are used to describe the processes and interactions at the microscopic level that lead to segregation, are analysed. Numerical simulations and comparisons with experimental data are then presented and conclusions are drawn on the capability of the model to accurately simulate the behaviour of granular matter during flow.