Data-driven intelligent optimisation of discontinuous composites

Abstract Fibre composites, and especially aligned discontinuous composites (ADCs), offer enormous versatility in composition, microstructure, and performance, but are difficult to optimise, due to their inherent variability and myriad permutations of microstructural design variables. This work combines an accurate yet efficient virtual testing framework (VTF) with a data-driven intelligent Bayesian optimisation routine, to maximise the mechanical performance of ADCs for a number of single- and multi-objective design cases. The use of a surrogate model helps to minimise the number of optimisation iterations, and provides a more accurate insight into the expected performance of materials which feature significant variability. Results from the single-objective optimisation study show that a wide range of structural properties can be achieved using ADCs, with a maximum stiffness of 505 GPa, maximum ultimate strain of 3.94%, or a maximum ultimate strength of 1.92 GPa all possible. A moderate trade-off in performance can be achieved when considering multi-objective optimisation design cases, such as an optimal ultimate strength & ultimate strain combination of 982 MPa and 3.27%, or an optimal combination of 720 MPa yield strength & 1.91% pseudo-ductile strain.

[1]  Fernando de la Prieta,et al.  Artificial neural networks used in optimization problems , 2018, Neurocomputing.

[2]  Jenn‐Ming Yang,et al.  Tensile properties of ultrahigh strength PAN-based, ultrahigh modulus pitch-based and high ductility pitch-based carbon fibers , 2008 .

[3]  A. Bismarck,et al.  Adhesion: Comparison Between Physico-chemical Expected and Measured Adhesion of Oxygen-plasma-treated Carbon Fibers and Polycarbonate , 2000 .

[4]  S. Pimenta,et al.  Semi-analytical simulation of aligned discontinuous composites , 2017 .

[5]  Jonas Mockus,et al.  Application of Bayesian approach to numerical methods of global and stochastic optimization , 1994, J. Glob. Optim..

[6]  Adam D. Bull,et al.  Convergence Rates of Efficient Global Optimization Algorithms , 2011, J. Mach. Learn. Res..

[7]  Donald R. Jones,et al.  A Taxonomy of Global Optimization Methods Based on Response Surfaces , 2001, J. Glob. Optim..

[8]  K. Potter,et al.  A novel manufacturing method for aligned discontinuous fibre composites (High Performance-Discontinuous Fibre method) , 2014 .

[9]  L. Gorbatikh,et al.  Hybrid effects in thin ply carbon/glass unidirectional laminates: Accurate experimental determination and prediction , 2016 .

[10]  M. Wisnom.  Mechanisms to create high performance pseudo-ductile composites , 2016 .

[11]  K. Pickering,et al.  Weak link scaling analysis of high-strength carbon fibre , 1999 .

[12]  H. Wu,et al.  Weibull analysis of strength-length relationships in single Nicalon SiC fibres , 1992 .

[13]  Andreas Krause,et al.  Information-Theoretic Regret Bounds for Gaussian Process Optimization in the Bandit Setting , 2009, IEEE Transactions on Information Theory.

[14]  S. Pimenta,et al.  Modelling hybrid effects on the stiffness of aligned discontinuous composites with hybrid fibre-types , 2017 .

[15]  G. Chiandussi,et al.  Comparison of multi-objective optimization methodologies for engineering applications , 2012, Comput. Math. Appl..

[16]  Y. Mai,et al.  On the single fibre pull-out problem: effect of loading method , 1992 .

[17]  Milo S. P. Shaffer,et al.  Exploring the pseudo-ductility of aligned hybrid discontinuous composites using controlled fibre-type arrangements , 2017 .

[18]  M. Z. Cohn,et al.  MULTIOBJECTIVE OPTIMIZATION OF PRESTRESSED CONCRETE STRUCTURES , 1993 .

[19]  A. F. Grandt,et al.  Fundamentals of Structural Integrity: Damage Tolerant Design and Nondestructive Evaluation , 2003 .

[20]  T. Rendall,et al.  Quasi-Isotropic and Pseudo-Ductile Highly Aligned Discontinuous Fibre Composites Manufactured with the HiPerDiF (High Performance Discontinuous Fibre) Technology , 2019, Materials.

[21]  S. Pimenta,et al.  Virtual testing framework for hybrid aligned discontinuous composites , 2017 .

[22]  Kalyanmoy Deb,et al.  Multi-objective Optimization , 2014 .

[23]  Tao Wang,et al.  Automatic Gait Optimization with Gaussian Process Regression , 2007, IJCAI.

[24]  K. Potter,et al.  Hierarchical pseudo-ductile hybrid composites combining continuous and highly aligned discontinuous fibres , 2018 .

[25]  Y. Mai,et al.  On steady-state fibre pull-outI The stress field , 1999 .

[26]  P. D. Mangalgiri.  Composite materials for aerospace applications , 1999 .

[27]  K. Potter,et al.  Aligned discontinuous intermingled reclaimed/virgin carbon fibre composites for high performance and pseudo-ductile behaviour in interlaminated carbon-glass hybrids , 2017 .

[28]  Cheng Li,et al.  Rapid Bayesian optimisation for synthesis of short polymer fiber materials , 2017, Scientific Reports.

[29]  Shuguang Li,et al.  Establishing size effects in discontinuous fibre composites using 2D finite element analysis , 2012 .

[30]  Milo S. P. Shaffer,et al.  The influence of variability and defects on the mechanical performance of tailorable composites , 2020, Journal of Composite Materials.

[31]  Zoubin Ghahramani,et al.  Dropout as a Bayesian Approximation: Representing Model Uncertainty in Deep Learning , 2015, ICML.

[32]  D. Edie,et al.  A comparative analysis of alternative models to predict the tensile strength of untreated and surface oxidised carbon fibers , 2001 .

[33]  R. Young,et al.  Effects of plasma oxidation on the surface and interfacial properties of ultra-high modulus carbon fibres , 2001 .

[34]  Mitsuo Gen,et al.  A Bayesian Optimization-based Evolutionary Algorithm for Flexible Job Shop Scheduling , 2015, Complex Adaptive Systems.

[35]  M. Wisnom,et al.  Pseudo-ductility and damage suppression in thin ply CFRP angle-ply laminates , 2015 .

[36]  Donald R. Jones,et al.  Efficient Global Optimization of Expensive Black-Box Functions , 1998, J. Glob. Optim..

[37]  David Amsallem,et al.  Automatised selection of load paths to construct reduced-order models in computational damage micromechanics: from dissipation-driven random selection to Bayesian optimization , 2016, Computational Mechanics.

[38]  Jasper Snoek,et al.  Practical Bayesian Optimization of Machine Learning Algorithms , 2012, NIPS.

[39]  Anil N. Netravali,et al.  Composites get greener , 2003 .

[40]  Carl E. Rasmussen,et al.  Gaussian processes for machine learning , 2005, Adaptive computation and machine learning.

[41]  Michael R Wisnom,et al.  Pseudo-ductility in intermingled carbon/glass hybrid composites with highly aligned discontinuous fibres , 2015 .

[42]  James K. Wessel,et al.  Handbook of advanced materials : enabling new designs , 2004 .