WP1: A simple to understand “superposition” approach (supported by advanced error estimation) had been implemented through fully integrated Abaqus plugins. Focus in the project is placed on non-flying panel assemblies (as these are easily distributed and inclusion in journal publications does not prevent dissemination).
WP2: Given the type of multiscale analysis, errors in macroscopic “driving” model degrees of freedom were identified as important. All error/uncertainty estimation methods developed within MARQUESS have looked to quantify and propagate this source through the multiscale problems in an computationally efficient way, which are limited to epistemic uncertainties in macroscopic (global) models. Therefore, uncertainty in composite microscopic (detailed feature) models were included. After literature review, it was decided that the most pressing area of uncertainty at this scale is the description of interlaminar conditions, particularly near the point of failure (delamination). Delamination modelling strategies were compared with experimental data and each other to determine the most appropriate approach for the superposition methods implemented in MARQUESS.
Goal oriented error estimation is the most appropriate uncertainty analysis tool for the MARQUESS approach as it furnishes the system with quantified estimates of driving degree of freedom uncertainties. A publicly available Python based system has been released which demonstrates the ZZ based GOEE method for shells through working examples.
WP3: A conceptually simple superposition multiscale modelling approach, and implemented in the final plugin deliverable and supports multiple relevant feature models (composite fillet sections, bonded joints, “sunken” fillets). The complete plugin incorporates feature model generation, “M” matrix population, and multiscale sub modelling. The plugin incorporates a user friendly GUI to improve adoption. Error estimation tools are demonstrated via a user editable Python based GOEE implementation.
WP4: Internal software development and unit testing was completed. These involved comparing the MARQUESS precomputed solution with standard linear sub modelling approaches. In all cases (following minor bug fixes) the models showed excellent agreement (with differences limited to 1-2%).
WP5: GOEE systems were designed to utilise the same HDF5 structure as the main MARQUESS workflow, thereby allowing excellent levels of compatibility between the two computational processes. Parametrised models for identified feature models (fillet section, curved fillet section, sunk fillet section, and bonded joint) were produced allowing for discrete lamina realisation using continuum elements, with full user control on layup sequence and overall component geometry.
