Machine learning-combined topology optimization for functionary graded composite structure design

Kim, Cheolwoong; Lee, Jaejoon; Yoo, Jeonghoon

doi:10.1016/j.cma.2021.114158

Cited by 44 publications

(10 citation statements)

References 69 publications

Supporting

Mentioning

Contrasting

Order By: Relevance

“…AI and related methods may have the potential to find correlations in the interaction between damage present in the FRP composites and the complex environmental and service load effects. This is not limited to material characterisation and analysis, AI may also be applied to structural design [139][140][141][142]. Such investigations yielding clear correlations, in the end, may yield insight into the underlying physical mechanisms and possibly contribute to reducing the experimental scatter, which are both required for robust and reliable predictive models and their validation.…”

Section: Perspectives For Cfrp Composite Structural Design Approaches...mentioning

confidence: 99%

In-Service Delaminations in FRP Structures under Operational Loading Conditions: Are Current Fracture Testing and Analysis on Coupons Sufficient for Capturing the Essential Effects for Reliable Predictions?

2022

View full text Add to dashboard Cite

Quasi-static or cyclic loading of an artificial starter crack in unidirectionally fibre-reinforced composite test coupons yields fracture mechanics data—the toughness or strain-energy release rate (labelled G)—for characterising delamination initiation and propagation. Thus far, the reproducibility of these tests is typically between 10 and 20%. However, differences in the size and possibly the shape, but also in the fibre lay-up, between test coupons and components or structures raise additional questions: Is G from a coupon test a suitable parameter for describing the behaviour of delaminations in composite structures? Can planar, two-dimensional, delamination propagation in composite plates or shells be properly predicted from essentially one-dimensional propagation in coupons? How does fibre bridging in unidirectionally reinforced test coupons relate to delamination propagation in multidirectional lay-ups of components and structures? How can multiple, localised delaminations—often created by impact in composite structures—and their interaction under service loads with constant or variable amplitudes be accounted for? Does planar delamination propagation depend on laminate thickness, thickness variation or the overall shape of the structure? How does exposure to different, variable service environments affect delamination initiation and propagation? Is the microscopic and mesoscopic morphology of FRP composite structures sufficiently understood for accurate predictive modelling and simulation of delamination behaviour? This contribution will examine selected issues and discuss the consequences for test development and analysis. The discussion indicates that current coupon testing and analysis are unlikely to provide the data for reliable long-term predictions of delamination behaviour in FRP composite structures. The attempts to make the building block design methodology for composite structures more efficient via combinations of experiments and related modelling look promising, but models require input data with low scatter and, even more importantly, insight into the physics of the microscopic damage processes yielding delamination initiation and propagation.

show abstract

Section: Perspectives For Cfrp Composite Structural Design Approaches...mentioning

confidence: 99%

In-Service Delaminations in FRP Structures under Operational Loading Conditions: Are Current Fracture Testing and Analysis on Coupons Sufficient for Capturing the Essential Effects for Reliable Predictions?

2022

View full text Add to dashboard Cite

show abstract

“…Machine learning, a distinct subset of artificial intelligence renowned for its capability to enable systems to learn from data and improve performance over time, has garnered substantial traction within the domain of additive manufacturing [12]. This advanced technology has proven to be a transformative force, driving innovation and optimization across various facets such as process parameter selection, design optimization, material selection, performance prediction, defect detection, and so on of the additive manufacturing process [13,14]. Additive manufacturing has been propelled into a new era of efficiency, precision, and customized production, underscoring the profound impact of machine learning within this context [15].…”

Section: Introductionmentioning

confidence: 99%

A machine learning-based recommendation framework for material extrusion fabricated triply periodic minimal surface lattice structures

Hussain,

Lee,

Tsang

et al. 2024

Preprint

View full text Add to dashboard Cite

Triply Periodic Minimal Surface (TPMS) lattice structures are utilized in diverse fields such as engineering, material design, and biomedical. The use of appropriate TPMS lattice structures in 3D printing can obtain benefits in terms of production efficiency and material reduction towards a greener 3D printing process. However, there is a lack of an automated solution to suggest the appropriate TPMS lattice structure parameters, such that unnecessary material wastage cannot be neglected in the existing practices. To address the above challenges, this study proposes a machine learning-based recommendation framework for generating the TPMS lattice structures based on the engineering requirements. First, we compiled a dataset by producing 144 samples via the material extrusion (ME) technique and conducted compression tests on four TPMS lattice structures (Diamond, Gyroid, Schwarz, and split-P), each with varying parameters, fabricated using Polylactic acid (PLA) material. Second, we train four machine learning algorithms (K-Nearest Neighbors, Decision Tree, Random Forest, and Bayesian Regression) on this dataset to predict TPMS lattice structure (unit cell type, unit cell size, and wall thickness). Extensive experiments assess algorithm performance using R-squared values and Root Mean Square Error (RMSE) as evaluation measures. Our results indicate that the Random Forest and Decision Tree algorithms perform best, achieving R-squared scores of 0.9694 and 0.9689, along with RMSE values of 0.1180 and 0.0795, respectively. This work not only advances the field's understanding of automated selection for TPMS lattice structures but also holds noteworthy implications for eco-design and eco-innovation, particularly in the realm of sustainable and efficient green 3D printing applications.

show abstract

“…Deep learning has some unique advantages in approximating nonlinear mappings of a data-based system. In turn, neural network models have been built to predict material constitutive relations [11,12], solve partial differential equations [13], solid mechanics [14][15][16], topology optimization [17][18][19], fracture problems [20][21][22], fluid flows [23], multiphysics problems in electrosurgery [24], finite element computations [25][26][27], and slope stability evaluation [28].…”

Section: Introductionmentioning

confidence: 99%

Modeling Geometrically Nonlinear FG Plates: A Fast and Accurate Alternative to IGA Method Based on Deep Learning

Li,

Yu,

Bui

2024

CMES

View full text Add to dashboard Cite

Isogeometric analysis (IGA) is known to show advanced features compared to traditional finite element approaches. Using IGA one may accurately obtain the geometrically nonlinear bending behavior of plates with functional grading (FG). However, the procedure is usually complex and often is time-consuming. We thus put forward a deep learning method to model the geometrically nonlinear bending behavior of FG plates, bypassing the complex IGA simulation process. A long bidirectional short-term memory (BLSTM) recurrent neural network is trained using the load and gradient index as inputs and the displacement responses as outputs. The nonlinear relationship between the outputs and the inputs is constructed using machine learning so that the displacements can be directly estimated by the deep learning network. To provide enough training data, we use S-FSDT Von-Karman IGA and obtain the displacement responses for different loads and gradient indexes. Results show that the recognition error is low, and demonstrate the feasibility of deep learning technique as a fast and accurate alternative to IGA for modeling the geometrically nonlinear bending behavior of FG plates.

show abstract

Machine learning-combined topology optimization for functionary graded composite structure design

Cited by 44 publications

References 69 publications

In-Service Delaminations in FRP Structures under Operational Loading Conditions: Are Current Fracture Testing and Analysis on Coupons Sufficient for Capturing the Essential Effects for Reliable Predictions?

In-Service Delaminations in FRP Structures under Operational Loading Conditions: Are Current Fracture Testing and Analysis on Coupons Sufficient for Capturing the Essential Effects for Reliable Predictions?

A machine learning-based recommendation framework for material extrusion fabricated triply periodic minimal surface lattice structures

Modeling Geometrically Nonlinear FG Plates: A Fast and Accurate Alternative to IGA Method Based on Deep Learning

Contact Info

Product

Resources

About