Computational Workflow Management in the context of Model-Based Design of Experiments

Abstract The project ”Modelling of morphology development of micro- and nanostructures” (MoDeNa) is developing a software framework specifically designed to facilitate functional dependencies in a multi-scale models using interconnected surrogate models; hence, an important component of the framework is to ensure that the surrogate models are valid. However, due to the functional dependencies between models, the model-based design of experiments (MBDoE) machine learning procedure, which itself is iterative, will inherently lead to the overall workflow becoming acyclic. This paper describes how the MoDeNa framework modularised the MBDoE procedure and embedded the modules into the computational workflow management tool FireWorks, thus ensuring that the framework supports acyclic workflows. A case study demonstrating how the principle works is also presented.