Abstract

The current industrial context is highly challenging. More agility is expected and reduced time to market strategies are key for businesses to grow. One key technology for advancing the product innovation process is virtual testing. An efficient implementation of virtual testing requires efficient and accurate material models. This is particularly true in the framework of complex materials like reinforced plastics and plastics for additive manufacturing. The classical workflow for generating virtual material laws is based on 3 main steps: First, costly experimental campaigns should be conducted. Second, post processing of the raw experimental data should be executed by dedicated human resources. Finally, the best suited material laws should be selected and implemented to be used in further virtual testing at part level. This process is labour intensive and time consuming. Consequently, it is not suited for today’s business pace. The proposed approach aims at enriching material databases by leveraging the power of experimental data, virtual testing and data science through a 4 steps workflow: First, an optimal experimental campaign is designed and executed. This first step is conducted smartly through a well-chosen Design of Experiments (DoE). The Smart DoE is designed using data science tools like correlation matrices, advanced sampling techniques and computation of reliability indices. Second, data science algorithms and tools are applied for filling smartly the gaps that are lacking in the initial spare experimental material database. The key step is based on combined data science tools like correlation, interpolation, proper order decomposition and reduced order modelling. Third, multiscale material modelling is deployed for connecting the different length scales. This is performed for different microstructures and different conditions. Fourth, a second data science pass is implemented aiming at enriching the final material database for different scenarios of microstructures and conditions. This is performed through data science tools like meta-modelling. The conducted work will report qualitative and quantitative results comparing the costing of classical approaches for building material databases to the proposed workflow. Considerable gains in time cost, data richness and overall Time to Market efficiency are reported and analysed.