Numerical and mathematical approaches for computation time reduction

This work, coordinated by UNIPI, was done with a direct contribution of all the Task 6.1 partners. This document provides first a review and a detailed analysis of existing models for Pellet-Cladding Mechanical Interaction (PCMI), the Fission Gas Behaviour (FGB) and Fuel Overfragmentation (FO) phenomena in industrial type fuel performance codes, such as TRANSURANUS, FINIX and CYRANO3, for which an improvement is expected by the substitution of an empirical formulation by machine-learning, surrogate modelling, or data-driven approaches. In a second part of the document a review is proposed for an identification and a presentation of the most powerful computation time reduction methods. The latter are decomposed in two categories with Machine Learning Methods (MLM) and surrogate models. In the last part of document some existing preliminary application of computation time reduction method for PCMI and FGB are discussed with more details. At the end of this review, we can conclude that the integration of MLM and surrogate models will bring significant progress in reducing computation time for the complex simulations needed for the fuel performance studies expected in the WP7 of the OperaHPC project. The implementation of the computation time reduction methods, specified in this document for PCMI, FGB and FO, can now start in the framework of the Tasks 6.2 and 6.3 of the WP6.