Browsing by Author "Millen, Scott L. J."

Now showing 1 - 3 of 3
  • Thumbnail Image
    ItemOpen Access
    Critical comparison of potential machine learning methods for lightning thermal damage assessment of composite laminates
    (Taylor and Francis, 2024-10-21) Lee, Juhyeong; Millen, Scott L. J.; Xu, Xiaodong
    The present study assesses the potential of using machine learning (ML) methods to predict the extent of lightning thermal damage in fiber-reinforced composite laminates using three supervised machine learning (SML) algorithms: (1) linear regression (LR), (2) decision tree (DT)-based, and (3) MLP models. These models were based on the 10 most significant factors that influence the severity of lightning damage, including three current waveform parameters, four material configurations, and three orthogonal electrical conductivities of each composite. All models demonstrated good performance with coefficient of determination (R2) values between 0.84 ~ 0.96. The multilayer perceptron (MLP) regression model trained with the lightning matrix damage dataset showed the most promising results (R2 > 0.94). Additional hyperparameter optimization was performed to improve the prediction performance of the baseline MLP model. The hyperparameter optimization (Adam optimizer, tanh activation function, and three hidden layers with 234 neurons) slightly improved the performance of the baseline MLP model by ~0.02, but achieved faster convergence. This result suggests that the baseline MLP model trained with the lightning matrix damage dataset is sufficiently accurate and robust. This paper highlights that ML-informed regression models can serve as an efficient first pass-estimator of lightning matrix damage in composite laminates, potentially reducing the amount of extremely time-consuming and expensive laboratory-scale lightning tests or streamlining the development of complex lightning damage models for future design.
  • Thumbnail Image
    ItemOpen Access
    On residual tensile strength after lightning strikes
    (Elsevier, 2025-07-01) Xu, Xiaodong; Millen, Scott L. J.; Mitchard, Daniel; Wisnom, Michael R.
    The study of post lightning strike residual strength is still relatively underdeveloped in the literature. Different approaches including in-plane compression or flexural testing have been used, but in-plane tensile loading post-strike has not been studied in detail. Although previous attempts have been made to determine the residual strength using Compression-After-Lightning (CAL) tests on composite laminates, these have been limited and not readily applicable under tensile loads. Therefore, this work completes Tension-After-Lightning (TAL) testing at 75 kA on composite laminates, a more realistic peak current than previously reported for TAL tests, to assess the knock-down in strength post-strike. The measured average TAL failure stress was 716 MPa, a reduction of 23 % from the baseline tensile failure stress of 929 MPa in the literature. This confirms a similar knock-down factor reported at lower peak currents (e.g. 50 kA), but the new TAL specimen geometry ensures that the lightning damage is contained within both the lightning and TAL specimen widths. In addition, a new Finite Element (FE) based virtual test was conducted, considering 0° ply splitting, and validated with the TAL tests herein. The TAL simulation predicted the residual tensile failure stress well, within 6 % of the measured value.
  • Thumbnail Image
    ItemOpen Access
    Towards a virtual test framework to predict residual compressive strength after lightning strikes
    (Elsevier, 2023-08-06) Millen, Scott L. J.; Xu, Xiaodong; Mukhopadhyay, Supratik; Wisnom, M. R.; Murphy, A.
    A novel integrated modelling framework is proposed as a set of coupled virtual tests to predict the residual compressive strength of carbon/epoxy composites after a lightning strike. Sequentially-coupled thermal-electric and thermo-mechanical models were combined with Compression After Lightning Strike (CAL) analyses, considering both thermal and mechanical lightning strike damage. The predicted lightning damage was validated using experimental images and X-ray Computed Tomography. Delamination and ply degradation information were mapped to a compression model, with a maximum stress criterion, using python scripts. Experimental data, in which artificial lightning strike and compression testing were performed, was used to assess the predictive capabilities of the framework, considering three lightning strike peak current amplitudes (25, 50, and 75 kA). The framework herein achieved a residual strength prediction within 6% of the experimental values for all peak currents. The relationship between individual lightning damage morphologies (thermal, mechanical and delamination damage) and CAL strength has been numerically established.