Digital Twin Modeling Method for Hierarchical Stiffened Plate Based on Transfer Learning

Xu, Ziyu; Gao, Tianhe; Li, Zengcong; Bi, Qingjie; Liu, Xiongwei; Tian, Kuo

doi:10.3390/aerospace10010066

Cited by 12 publications

(4 citation statements)

References 30 publications

Supporting

Mentioning

Contrasting

Order By: Relevance

“…Tian et al [27] proposed a transfer-learning-based variable-fidelity surrogate model (TL-VFM) for shell bulking prediction, which employs a two-stage training process to train deep neural networks (DNN) with multi-fidelity data. Xu et al [28] employed the TL-VFM to develop a digital twin for a hierarchical stiffened plate. Meng et al [13] proposed multifidelity deep neural networks (mDNNs) learning from multi-fidelity data to solve function approximation and inverse partial differential equation problems.…”

Section: Introductionmentioning

confidence: 99%

Extended Hierarchical Kriging Method for Aerodynamic Model Generation Incorporating Multiple Low-Fidelity Datasets

Pham,

Tyan,

Nguyen

et al. 2023

Aerospace

View full text Add to dashboard Cite

Multi-fidelity surrogate modeling (MFSM) methods are gaining recognition for their effectiveness in addressing simulation-based design challenges. Prior approaches have typically relied on recursive techniques, combining a limited number of high-fidelity (HF) samples with multiple low-fidelity (LF) datasets structured in hierarchical levels to generate a precise HF approximation model. However, challenges arise when dealing with non-level LF datasets, where the fidelity levels of LF models are indistinguishable across the design space. In such cases, conventional methods employing recursive frameworks may lead to inefficient LF dataset utilization and substantial computational costs. To address these challenges, this work proposes the extended hierarchical Kriging (EHK) method, designed to simultaneously incorporate multiple non-level LF datasets for improved HF model construction, regardless of minor differences in fidelity levels. This method leverages a unique Bayesian-based MFSM framework, simultaneously combining non-level LF models using scaling factors to construct a global trend model. During model processing, unknown scaling factors are implicitly estimated through hyperparameter optimization, resulting in minimal computational costs during model processing, regardless of the number of LF datasets integrated, while maintaining the necessary accuracy in the resulting HF model. The advantages of the proposed EHK method are validated against state-of-the-art MFSM methods through various analytical examples and an engineering case study involving the construction of an aerodynamic database for the KP-2 eVTOL aircraft under various flying conditions. The results demonstrated the superiority of the proposed method in terms of computational cost and accuracy when generating aerodynamic models from the given multi-fidelity datasets.

show abstract

Section: Introductionmentioning

confidence: 99%

Extended Hierarchical Kriging Method for Aerodynamic Model Generation Incorporating Multiple Low-Fidelity Datasets

Pham,

Tyan,

Nguyen

et al. 2023

Aerospace

View full text Add to dashboard Cite

show abstract

“…Effectively, a digital twin is constructed whose goal is to faithfully replicate the physical components of the test-rig and output structural data comparable to the experimentally derived sensor data. In the work laid out by Xu et al (2023), a digital twin model for strength health monitoring of stiffened plates for use in the aerospace sector has been built after being trained with a deep neural network algorithm. An alternative approach is selected in this paper, which uses detailed simulations of physical tests of both the “fuselage barrel” and the “stiffened panel”.…”

Section: Introductionmentioning

confidence: 99%

Design and development of an experimental bench concept for testing of fuselage stiffened panels, using a virtual testing methodology

Kordas,

Fotopoulos,

Lampeas

et al. 2023

AEAT

View full text Add to dashboard Cite

Purpose Fuselage structures are subjected to combinations of axial, bending, shear and differential pressure loads. The validation of advanced metallic and composite fuselage designs against such loads is based on the full-scale testing of the fuselage barrel, which, however, is highly demanding from a time and cost viewpoint. This paper aims to assist in scaling-down the experimentation to the stiffened panel level which presents the opportunity to validate state-of-the-art designs at higher rates than previously attainable. Design/methodology/approach Development of a methodology to successfully design tests at the stiffened panel level and realize them using advanced, complex and adaptable test-rigs that are capable of introducing independently a set of distinct load types (e.g. internal overpressure, tension, shear) while applying appropriate boundary conditions at the edges of the stiffened panel. Findings A baseline test-rig configuration was developed after extensive parametric modelling studies at the stiffened panel level. The realization of the loading and boundary conditions on the test-rig was facilitated through innovative supporting and loading system set-ups. Originality/value The proposed test bench is novel and compared to the conventional counterparts more viable from an economic and manufacturing point of view. It leads to panel responses, which are as close as possible to those of the fuselage barrel in-flight and can be used for the execution of static or fatigue tests on metallic and thermoplastic curved integrally stiffened full-scale panels, representative of a business jet fuselage.

show abstract

“…Gomes et al [8] proposed a damage identification method on a helicopter's main rotor blade by combining a FE model with a bat optimization algorithm to An ANN for damage detection, trained using in situ distributed strained have been implemented by Califano et al [9]. Xu et al [10] developed a DT modeling method for hierarchical stiffened panel. Based on a pre-trained deep neural network, the model has been fine-tuned and finally, the DT was capable to visualize the full-field strength of the structure.…”

Section: Fan Blade Bird Strike 1 Introductionmentioning

confidence: 99%

A Digital Twin of a Fiber Bragg Grating Network in the Leading Edge of a Rotating Composite Fan Blade Subjected to Bird Strike

Floros,

Sotiropoulos,

Tserpes

et al. 2023

10th ECCOMAS Thematic Conference on Smart Structures and Materials

View full text Add to dashboard Cite

In the last two decades, an intensive effort is in being made to manufacture smart composite structures with embedded fiber optic sensors that can be exploited to detect, localize, and classify damage. In this context, the use of digital twins for the virtual design and optimization of a sensors' network may give insight to complex phenomena and significantly reduce the development time and cost of the smart structures. Composite engine fan blades are prone to bird strike damage and thus, early detection of bird strike damage is crucial for the efficient maintenance of the engines and the prevention of critical incidents. In this work, a Fiber Bragg Grating (FBG) network in the leading edge of a rotating composite fan blade subjected to bird strike is developed. The fan blade is made of 3D woven composite material and its leading edge is protected by an adhesively bonded layer made of steel while the FBG network is placed into the bondline. The Digital Twin (DT) of the above description integrates a Finite Element (FE) model and Artificial Neural Network (ANN) models. In the FE model, progressive damage in the composite material and debonding of the steel layer have been simulated. Several bird strike simulation scenarios were performed at different strike locations. At the time of the bird strike, the blade is rotating with a rotation speed of 3000 rpm. ANNs were trained using the numerical strain data in order to predict the existence of debonding of the steel layer and the area of debonding. ANNs exploit Convolutional Neural Networks (CNNs) to capture spatial dependencies in input signals. Experimental results show ANNs provide low root mean square error (RMSE) in both regression tasks, effectively predicting the bird strike location and the damage in terms of the relevant affecting area.

show abstract

Digital Twin Modeling Method for Hierarchical Stiffened Plate Based on Transfer Learning

Cited by 12 publications

References 30 publications

Extended Hierarchical Kriging Method for Aerodynamic Model Generation Incorporating Multiple Low-Fidelity Datasets

Extended Hierarchical Kriging Method for Aerodynamic Model Generation Incorporating Multiple Low-Fidelity Datasets

Design and development of an experimental bench concept for testing of fuselage stiffened panels, using a virtual testing methodology

A Digital Twin of a Fiber Bragg Grating Network in the Leading Edge of a Rotating Composite Fan Blade Subjected to Bird Strike

Contact Info

Product

Resources

About