Surrogate-based multiscale analysis of experiments on thermoplastic composites under off-axis loading

TL;DR

The study tackles the challenge of predicting macroscale responses of off-axis UD thermoplastic composites by exposing the breakdown of macroscopic homogeneity at low off-axis angles. It introduces a surrogate-based multiscale framework that replaces the computationally expensive micromodel with a Physically Recurrent Neural Network (PRNN) embedded with constitutive models, combined with transfer-learning in the latent space to extrapolate material properties without retraining. The authors demonstrate improved agreement with constant-strain-rate experiments across a range of off-axis angles and elucidate the macroscopic strain-field inhomogeneity, while also exploring oblique end-tabs to reduce stress concentrations; they further extend the methodology to creep tests, revealing limitations tied to machine grip calibration and the need for dedicated spectrum training. Overall, the approach provides a computationally feasible pathway to robust multiscale analysis and design guidance for thermoplastic composites under complex loading, with practical implications for experimental setups and material-property extrapolation.

Abstract

In this paper, we present a surrogate-based multiscale approach to model constant strain-rate and creep experiments on unidirectional thermoplastic composites under off-axis loading. In previous contributions, these experiments were modeled through a single-scale micromechanical simulation under the assumption of macroscopic homogeneity. Although efficient and accurate in many scenarios, simulations with low-off axis angles showed significant discrepancies with the experiments. It was hypothesized that the mismatch was caused by macroscopic inhomogeneity, which would require a multiscale approach to capture it. However, full-field multiscale simulations remain computationally prohibitive. To address this issue, we replace the micromodel with a Physically Recurrent Neural Network (PRNN), a surrogate model that combines data-driven components with embedded constitutive models to capture history-dependent behavior naturally. The explainability of the latent space of this network is also explored in a transfer learning strategy that requires no re-training. With the surrogate-based simulations, we confirm the hypothesis raised on the inhomogeneity of the macroscopic strain field and gain insights into the influence of adjustment of the experimental setup with oblique end-tabs. Results from the surrogate-based multiscale approach show better agreement with experiments than the single-scale micromechanical approach over a wide range of settings, although with limited accuracy on the creep experiments, where macroscopic test effects were implicitly taken into account in the material properties calibration.

Figures (19)

• Figure 1: Schematic representation of UD carbon/PEEK composite system specimen and creep testing setup.
• Figure 2: Scheme of concurrent multiscale framework with two scales (macro $\Omega$ and micro $\omega$) for composite material with off-axis loading.
• Figure 3: Alternative approaches to replace full multiscale problem for modeling composites under off-axis loading.
• Figure 4: Architecture of PRNN and fictitious material point unrolled in time.
• Figure 5: From training with micromodel with a set of properties $\boldsymbol{\theta}_{(\cdot)}^{\textrm{A}}$ to extrapolating to micromodel with set of properties $\boldsymbol{\theta}_{(\cdot)}^{\textrm{B}}$ without retraining.