Abstract
In this work, the authors present a digital twin (DT) to analyse the manufacturing process of structural composites by using resin transfer moulding (RTM). During RTM, a dry textile preform is impregnated with a polymer resin injected in a closed mould. RTM is one of the most used production methods for high-performance structural composites. The DT is focused on detecting in-homogeneous resin flow produced by race-tracking channels that divert resin flow to the outlet gates of the mould, producing dry spots and lack of impregnation. The DT core contains two surrogate models based on encoder/decoder deep learning architectures, providing the fast/accurate response necessary for interrogation during manufacturing. The first surrogate acts as the disturbance detector, providing the on-the-fly representation of the fabric permeability with the only information gathered by a set of five pressure sensors distributed over the mould surface. The second offers real-time representation of a set of quantities of interests (QoI): namely, the flow progress and the pressure field inside the mould. Training of both surrogates was performed with synthetic data generated by high-fidelity multi-physics simulations of the flow progress in a porous preform by following Darcy’s law. Errors in the pressure field predictions of the surrogates are lower than 1\(\%\) with consultation time <50 ms, enabling encapsulation in the digital twin. The DT performance was evaluated by comparing the response against a set of RTM experiments for different race-tracking scenarios. The two most relevant novelties of the work are the use of the concept of the instantiated DT, which provides information on the current state of the process from data provided by a network of distributed sensors, and that this DT has been trained exclusively with synthetic data from multiphysics simulations while evaluated against experimental data from injection tests.
Similar content being viewed by others
References
Barron JT (2019) A general and adaptive robust loss function. In: Proceedings of the IEEE conference on computer vision and pattern recognition, pp 4331–4339
Bickerton S, Advani SG (1999a) Characterization and modeling of race-tracking in liquid composite molding processes. Compos Sci Technol 59(15):2215–2229. https://doi.org/10.1016/S0266-3538(99)00077-9, https://www.sciencedirect.com/science/article/pii/S0266353899000779
Bickerton S, Advani SG (1999b) Characterization and modeling of race-tracking in liquid composite molding processes. Compos Sci Technol 59(15):2215–2229. https://doi.org/10.1016/S0266-3538(99)00077-9, https://www.sciencedirect.com/science/article/pii/S0266353899000779
Brovkova M, Molodtsov V, Bushuev V (2021) Implementation specifics and application potential of digital twins of technological systems. The International Journal of Advanced Manufacturing Technology 117(7):2279–2286. https://doi.org/10.1007/s00170-021-07141-z
BWang X, Bo L, Fuxin L (2019) Adaptive wing loss for robust face alignment via heatmap regression. In: Proceedings of the international conference on computer vision, pp 6971–6981
Caglar B, Salvatori D, Sozer EM et al (2018) In-plane permeability distribution mapping of isotropic mats using flow front detection. Compos A Appl Sci Manuf 113:275–286. https://doi.org/10.1016/j.compositesa.2018.07.036, https://www.sciencedirect.com/science/article/pii/S1359835X18303063
Canal LP, Benavente M, Hausmann M et al (2015) Process-induced strains in RTM processing of polyurethane/carbon composites. Compos A Appl Sci Manuf 78:264–273. https://doi.org/10.1016/j.compositesa.2015.08.018, https://www.sciencedirect.com/science/article/pii/S1359835X15002894
Crawford B, Rashif KMS, Rashidi A et al (2020) A Bayesian belief approach to quality control of resin transfer molding process. The International Journal of Advanced Manufacturing Technology 109(7):1953–1968. https://doi.org/10.1007/s00170-020-05715-x
Devillard M, Hsiao KT, Gokce A et al (2003a) On-line characterization of bulk permeability and race-tracking during the filling stage in resin transfer molding process. J Compos Mater 37(17):1525–1541
Devillard M, Hsiao KT, Gokce A et al (2003b) On-line characterization of bulk permeability and race-tracking during the filling stage in resin transfer molding process. J Compos Mater 37(17):1525–1541. https://doi.org/10.1177/0021998303034459
Di Fratta C, Klunker F, Ermanni P (2013) A methodology for flow-front estimation in LCM processes based on pressure sensors. Compos A Appl Sci Manuf 47:1–11. https://doi.org/10.1016/j.compositesa.2012.11.008, https://www.sciencedirect.com/science/article/pii/S1359835X12003429
Feng ZH, Kittler J, Awais M et al (2017) Wing loss for robust facial landmark localisation with convolutional neural networks. https://doi.org/10.48550/ARXIV.1711.06753, arXiv:1711.06753
Feng ZH, Kittler J, Awais M et al (2020) Rectified wing loss for efficient and robust facial landmark localisation with convolutional neural networks. Int J Comput Vision 128(8):2126–2145. https://doi.org/10.1007/s11263-019-01275-0
Fernández-León J, Keramati K, Garoz D et al (2022) A machine learning strategy for race-tracking detection during manufacturing of composites by liquid moulding. Integrating Materials and Manufacturing Innovation 11(2):296–311. https://doi.org/10.1007/s40192-022-00263-6
Fernández-León J, Keramati K, Miguel C et al (2023) A deep encoder-decoder for surrogate modelling of liquid moulding of composites. Eng Appl Artif Intell 120:105,945. https://doi.org/10.1016/j.engappai.2023.105945, https://www.sciencedirect.com/ science/article/pii/S095219762300129X
Fratta CD, Koutsoukis G, Klunker F et al (2016) Fast method to monitor the flow front and control injection parameters in resin transfer molding using pressure sensors. J Compos Mater 50(21):2941–2957. https://doi.org/10.1177/0021998315614994
Friederich J, Francis DP, Lazarova-Molnar S et al (2022) A framework for data-driven digital twins of smart manufacturing systems. Comput Ind 136:103,586. https://doi.org/10.1016/j.compind.2021.103586, https://www.sciencedirect.com/science/ article/pii/S0166361521001937
González C, Vilatela J, Molina-Aldareguía J et al (2017) Structural composites for multifunctional applications: current challenges and future trends. Prog Mater Sci 89:194–251
Gunasegaram D, Murphy A, Barnard A et al (2021a) Towards developing multiscale-multiphysics models and their surrogates for digital twins of metal additive manufacturing. Addit Manuf 46:102,089. https://doi.org/10.1016/j.addma.2021.102089, https://www.sciencedirect.com/science/article/pii/S2214860421002542
Gunasegaram DR, Murphy AB, Matthews MJ et al (2021b) The case for digital twins in metal additive manufacturing. Journal of Physics: Materials 4(4):040,401. https://doi.org/10.1088/2515-7639/ac09fb
Heo TW, Khairallah SA, Shi R et al (2021) A mesoscopic digital twin that bridges length and time scales for control of additively manufactured metal microstructures. Journal of Physics: Materials 4(3):034,012. https://doi.org/10.1088/2515-7639/abeef8
Huang G, Liu Z, Van Der Maaten L et al (2017) Densely connected convolutional networks. In: Proceedings of the IEEE conference on computer vision and pattern recognition, pp 4700–4708
Karpatne A, Atluri G, Faghmous JH et al (2017) Theory-guided data science: a new paradigm for scientific discovery from data. IEEE Trans Knowl Data Eng 29(10):2318–2331. https://doi.org/10.1109/TKDE.2017.2720168
Kumar S, Gopi T, Harikeerthana N et al (2023) Machine learning techniques in additive manufacturing: a state of the art review on design, processes and production control. J Intell Manuf 34(1):21–55. https://doi.org/10.1007/s10845-022-02029-5
Lekakou C, Cook S, Deng Y et al (2006) Optical fibre sensor for monitoring flow and resin curing in composites manufacturing. Compos A Appl Sci Manuf 37(6):934–938. https://doi.org/10.1016/j.compositesa.2005.03.003, https://www.sciencedirect.com/science/article/pii/S1359835X05000965, selected Contributions from the 7th international conference on flow processes in composite materials held at University of Delaware, USA
Liu J, Guo F, Gao H et al (2021) Defect detection of injection molding products on small datasets using transfer learning. J Manuf Process 70:400–413. https://doi.org/10.1016/j.jmapro.2021.08.034, https://www.sciencedirect.com/science/article/pii/S1526612521006113
LLorca J, González C, Molina-Aldareguía JM et al (2011) Multiscale modeling of composite materials: a roadmap towards virtual testing. Adv Mater 23(44):5130–5147
Modoni GE, Stampone B, Trotta G (2022) Application of the digital twin for in process monitoring of the micro injection moulding process quality. Comput Ind 135:103,568. https://doi.org/10.1016/j.compind.2021.103568, https://www.sciencedirect.com/science/ article/pii/S0166361521001755
Ogoke F, Farimani AB (2021) Thermal control of laser powder bed fusion using deep reinforcement learning. Addit Manuf 46:102,033. https://doi.org/10.1016/j.addma.2021.102033, https://www.sciencedirect.com/science/article/pii/S2214860421001986
Paszke A, Gross S, Massa F et al (2019) PyTorch: an imperative style, high-performance deep learning library. In: Wallach H, Larochelle H, Beygelzimer A et al (eds) Advances in neural information processing systems 32. Curran Associates, Inc., p 8024–8035, http://papers.neurips.cc/paper/9015-pytorch-an-imperative-style-high-performance-deep-learning-library.pdf
Peng GCY, Alber M, Buganza Tepole A et al (2021) Multiscale modeling meets machine learning: what can we learn? Archives of Computational Methods in Engineering 28(3):1017–1037. https://doi.org/10.1007/s11831-020-09405-5
Phua A, Davies C, Delaney G (2022) A digital twin hierarchy for metal additive manufacturing. Comput Ind 140:103,667. https://doi.org/10.1016/j.compind.2022.103667, https://www.sciencedirect.com/science/article/pii/S0166361522000641
Redelinghuys AJH, Basson AH, Kruger K (2020) A six-layer architecture for the digital twin: a manufacturing case study implementation. J Intell Manuf 31(6):1383–1402. https://doi.org/10.1007/s10845-019-01516-6
Rosen R, von Wichert G, Lo G et al (2015) About the importance of autonomy and digital twins for the future of manufacturing. IFAC-PapersOnLine 48(3):567–572. https://doi.org/10.1016/j.ifacol.2015.06.141, https://www.sciencedirect.com/science/article/pii/S2405896315003808, 15th IFAC Symposium on information control problems in manufacturing
Rossi A, Moretti M, Senin N (2022) Neural networks and NARXs to replicate extrusion simulation in digital twins for fused filament fabrication. J Manuf Process 84:64–76. https://doi.org/10.1016/j.jmapro.2022.09.048, https://www.sciencedirect.com/science/article/pii/S1526612522006685
Sánchez del Río J, Pascual-González C, Martínez V et al (2021) 3D-printed resistive carbon-fiber-reinforced sensors for monitoring the resin frontal flow during composite manufacturing. Sens Actuators, A 317:112,422. https://doi.org/10.1016/j.sna.2020.112422, https://www.sciencedirect.com/science/article/pii/S0924424720317386
Schleich B, Anwer N, Mathieu L et al (2017) Shaping the digital twin for design and production engineering. CIRP Ann 66(1):141–144. https://doi.org/10.1016/j.cirp.2017.04.040, https://www.sciencedirect.com/science/article/pii/S0007850617300409
Skordos AA, Karkanas PI, Partridge IK (2000) A dielectric sensor for measuring flow in resin transfer moulding. Meas Sci Technol 11(1):25. https://doi.org/10.1088/0957-0233/11/1/304
Stieber S, Schröter N, Schiendorfer A et al (2021) FlowFrontNet: improving carbon composite manufacturing with CNNs. In: Dong Y, Mladenić D, Saunders C (eds) Machine learning and knowledge discovery in databases: applied data science track. Springer International Publishing, Cham, pp 411–426
Szarski M, Chauhan S (2023) Instant flow distribution network optimization in liquid composite molding using deep reinforcement learning. J Intell Manuf 34(1):197–218. https://doi.org/10.1007/s10845-022-01990-5
Tao F, Qi Q, Wang L et al (2019) Digital twins and cyber-physical systems toward smart manufacturing and Industry 4.0: correlation and comparison. Engineering 5(4):653–661. https://doi.org/10.1016/j.eng.2019.01.014, https://www.sciencedirect.com/science/article/pii/S209580991830612X
Tercan H, Meisen T (2022) Machine learning and deep learning based predictive quality in manufacturing: a systematic review. J Intell Manuf 33(7):1879–1905. https://doi.org/10.1007/s10845-022-01963-8
Wang J, Ma Y, Zhang L et al (2018) Deep learning for smart manufacturing: methods and applications. J Manuf Syst 48:144–156. https://doi.org/10.1016/j.jmsy.2018.01.003, https://www.sciencedirect.com/science/article/pii/S0278612518300037, special Issue on smart manufacturing
Wang Z, Yang W, Liu Q et al (2022) Data-driven modeling of process, structure and property in additive manufacturing: a review and future directions. J Manuf Process 77:13–31. https://doi.org/10.1016/j.jmapro.2022.02.053, https://www.sciencedirect.com/science/article/pii/S1526612522001529
Weller HG, Tabor G, Jasak H et al (1998) A tensorial approach to computational continuum mechanics using object-oriented techniques. Comput Phys 12(6):620–631. https://doi.org/10.1063/1.168744
Wuest T, Weimer D, Irgens C et al (2016) Machine learning in manufacturing: advantages, challenges, and applications. Production & Manufacturing Research 4(1):23–45. https://doi.org/10.1080/21693277.2016.1192517
Yavari R, Riensche A, Tekerek E et al (2021) Digitally twinned additive manufacturing: detecting flaws in laser powder bed fusion by combining thermal simulations with in-situ meltpool sensor data. Mater Des 211:110,167. https://doi.org/10.1016/j.matdes.2021.110167, https://www.sciencedirect.com/science/article/pii/S026412752100722X
Zhang R, Wang F, Cai J et al (2022) Digital twin and its applications: a survey. The International Journal of Advanced Manufacturing Technology 123(11):4123–4136. https://doi.org/10.1007/s00170-022-10445-3
Funding
The research leading to the results presented in this paper received funding from the Regional Government of Madrid through the research and innovation Innovation Hubs 2018 through the TEMACOM project (49.520635.9.18) and MAT4.0-CM (S2018/NMT-4381) projects. Luis Baumela is part of Ellis Unit Madrid, research partially funded by the Autonomous Community of Madrid. The authors would like to thank Dr. Davide Mocerino, Dr. David Garoz and Mr. Jon Urtasun for their useful discussions and input throughout the work. They also gratefully acknowledge the support given by the NVIDIA Academic Hardware Grants Program. The authors gratefully acknowledge the Universidad Politécnica de Madrid for providing computing resources on the Magerit Supercomputer.
Author information
Authors and Affiliations
Contributions
Joaquín Fernández-León, Luis Baumela and Carlos González designed the original idea of the digital twin model. Keayvan Keramati carried out the experimental injection tests. Joaquín Fernández-León implemented the model and performed the correlation against the experiments. Luis Baumela and Carlos González conceived the study and oversaw overall planning. All authors discussed the results and contributed to the final manuscript.
Corresponding author
Ethics declarations
Competing interests
The authors declare no competing interests.
Additional information
Publisher's Note
Springer Nature remains neutral with regard to jurisdictional claims in published maps and institutional affiliations.
Supplementary Information
Below is the link to the electronic supplementary material.
Appendix A: Composite loss function
Appendix A: Composite loss function
Training is carried out by minimizing the loss function defined as
where \(r_i\) stands for the residual (difference between the ground truth and the model prediction) and \(\rho (r_i)\) the error norm. To improve the accuracy of the model, it is necessary to encourage the identification of the regions in the model with large residuals. Inspired by recent results [1, 5, 12, 13], an error loss based on a wing shape [12, 13] that is of \(C^1\) differentiability class was introduced. Mathematically, \(\rho _{n-log}\)
with \(\delta _1\) the residual threshold, n an even natural number and \(C_1=1/(n \delta _1^{n-1}(1+\delta _1))\) and \(C_2=\ln (1+\delta _1)-C_1\delta _1^n\) parameters to ensure \(C^1\) continuity at \(r_i=\delta _1\). The \(n=4\) potential is introduced to identify differentiability at \(r_i=0\), while the natural logarithm is used to penalise the presence of large outliers. The factor \(\omega \) is used to increase the influence of residuals. The factors \(\delta _1=10^{-3}\) and \(\omega =2.0\) were selected in for optimum results [15] for optimum results, see Fig. 17 .
Rights and permissions
Springer Nature or its licensor (e.g. a society or other partner) holds exclusive rights to this article under a publishing agreement with the author(s) or other rightsholder(s); author self-archiving of the accepted manuscript version of this article is solely governed by the terms of such publishing agreement and applicable law.
About this article
Cite this article
Fernández-León, J., Keramati, K., Baumela, L. et al. A digital twin for smart manufacturing of structural composites by liquid moulding. Int J Adv Manuf Technol 130, 4679–4697 (2024). https://doi.org/10.1007/s00170-023-12637-x
Received:
Accepted:
Published:
Issue Date:
DOI: https://doi.org/10.1007/s00170-023-12637-x