(669b) Plant-Level Modeling and Simulation of Fuel Recycling Processes: Commonality and Integration | AIChE

(669b) Plant-Level Modeling and Simulation of Fuel Recycling Processes: Commonality and Integration


de Almeida, V. F. - Presenter, Oak Ridge National Laboratory

Plant-Level Modeling and Simulation of Fuel Recycling Processes: Commonality and Integration

Valmor F. de Almeida

Oak Ridge National Laboratory, Oak Ridge, TN 37831-6181


Used nuclear fuel recycling processes are known to be diverse and complex. This has led to a case-by-case approach for developing mathematical models and their corresponding simulation modules. This approach has many drawbacks. First, it does not provide a common theoretical framework to be applied to all processes at a well defined length and time scale. Second, as a result, it is difficult to couple mathematical models of different processes in the plant flowsheet and define a verification/validation strategy. Last, the resulting computational modules that reflect the implementation of these models, inherit the same difficulties at the algorithm and software engineering level. That is, the software realization of mathematical models are not intended to be integrated and lack a theoretical basis for coupling. In addition, computational modules do not take advantage of internal algorithms/solvers, and modules either do not have or reuse a software architecture. Given this status, modeling and simulation for fuel recycling processes at the plant level can take advantage of a modern approach taking into account commonality and integration as a basis for future developments. A desirable advantage of this approach is to provide a standard for verification, validation, and uncertainty quantification of models and modules of processes while reducing the cost of development. This scientific paradigm is needed in the nuclear recycling domain to improve robustness of simulation predictions, and gain public and political acceptance; the latter is a unique aspect of the nuclear recycling enterprise.

The work in this presentation is aimed at identifying a modeling commonality for various processes in a recycling plant using a network approach. A plant or flowsheet is modeled as a network of computational modules. These modules are designed with common external and internal behaviors to allow for integration and modular plant design. Internally, the modules result from the implementation of a model stemming from a physico-chemical network transport theory. The application of a common theory for all processes enables the development of a systematic set of modules. A central concept is the abstraction of contactors as a network of stages representing the mechanics of phase contact. This abstraction enables the development of a common structure for various separation processes wherein the application of equilibrium and non-equilibrium considerations will be independent of the particular contacting equipment. This approach allows for different kinetics and thermodynamics models to be applied to a contactor network without internal changes in the contactor model. In addition, the approach supports the development of a reduced-order model wherein subscale phenomena can be taken into account in a more systematic way. The resulting balance equations for the state variables of the network model provide the means for evolving the state of the system and for predicting quantities of practical interest. Time-coupling of modules is achieved in asynchronous way to allow for greater flexibility in developing models and integrating the resulting modules. This approach also leverages computation concurrency in parallel machines.

Computational examples of commonality and integration will be described for three processes of interest to aqueous recycling, namely, voloxidation, dissolution, and solvent extraction. These seemingly different processes share the above-mentioned commonalities and can be integrated in a more straightforward way than if they had been developed in isolation. A description of how model fidelity can be improved will be made by identifying quantities which can be provided by subscale models.