High-fidelity Finite Element Model Research Articles

Cyclic Behavior of Concrete-Filled Tube Columns with Bidirectional Moment Connections Considering the Local Slenderness Effect

In this research, the cyclic behavior of concrete-filled thin tube (CFTT) columns with bidirectional moment connections was numerically studied within the context of thin-walled structures. Novel considerations in the design of CFTT columns with slenderness sections are proposed through a parametric study. A total of 70 high-fidelity finite element (FE) models are developed using ANSYS software v2022 calibrated from experimental research using similar 3D joint configurations. Furthermore, a comparison of different width-to-thickness ratios in columns was considered. The results showed that the models with a high slenderness ratio reached a stable cyclic behavior until 0.03 rad of drift, and a flexural strength of 0.8 Mp was reached for 4% of the drift ratio according to the Seismic Provisions. However, this effect slightly decreased the strength and the dissipated energy of the moment connection in comparison to columns with a high ductility ratio. Moreover, an evaluation of concrete damages shows concrete cracked for cyclic loads higher than 3% of drift. Finally, the joint configurations studied can achieve a good performance, avoiding brittle failure mechanisms and ensuring the plastic hinges in the beams.

Computational engine for finite element digital twins of structural dynamics via motion data

Typical structural health monitoring systems employ limited numbers of sensors capable of measuring discrete local behaviours. However, practical challenges arise as these sensor arrays cannot cover all local areas of interest. To address this challenge, this article introduces a novel method for twinning structural dynamic behaviour by constructing a finite-element-model-based digital twin, enabling the observation of non-sensor positions crucial for downstream tasks. The approach utilises streaming monitoring data, e.g., displacement and acceleration, as external dynamic loads to reproduce the dynamic response of the entire physical structure. Subsequently, the dynamic behaviour of specific non-sensor locations can be extracted from the digital twin. The method is formulated as a local-global-local procedure. To validate the proposed approach, two virtual experiments were conducted on: 1) a simply supported Euler-Bernoulli beam subjected to static loads and 2) a high-fidelity finite element model of a composite bridge carrying dynamic traffic loads. The results demonstrate remarkable accuracy in reproducing both global and local behaviours, facilitating observations at non-sensor positions for downstream estimations.

Spatiotemporally nonlocal homogenization method for viscoelastic porous metamaterial structures

The underlying size-dependent mechanism of viscoelastic porous metamaterial structures is still unclear, and the solution to their response using the high-fidelity numerical method is time-consuming, even prohibitive. This paper explores how the size-dependent effect of metamaterial structures emerges from the underlying nonlocal interactions in metamaterial microstructures, and then develops an accurate but efficient homogenization method for analyzing their size-dependent and time-dependent mechanical behaviors of viscoelastic porous metamaterial structures. By accounting for the nonlocal interaction between different representative volume elements and the history-dependent effect, a spatiotemporally nonlocal discrete element model is developed to analyze the mechanical behavior of viscoelastic porous metamaterial structures. With the help of the spatiotemporally nonlocal discrete element model, a spatiotemporally nonlocal homogenization method is proposed to accurately yet efficiently assess the size-dependent mechanical behavior of the porous metamaterial structure. The remarkable potential of the spatiotemporally nonlocal homogenization method is illustrated by simulating the stress relaxation of a viscoelastic porous metamaterial structure under axial strain. Results indicate that the size-dependent phenomenon can be significant, and the spatiotemporally nonlocal homogenization model can efficiently capture the size-dependent relaxation stress evaluated by the high-fidelity finite element model without loss of accuracy.

Hygrothermal effects on ballistic behavior of toughened CFRP laminates

High-speed impact resistance is a critical consideration in the design and assessment of composites for aerospace applications during their service life, especially in the context of hygrothermal aging. This study, combining experiment and numerical simulation, investigates hygrothermal effects on the high-speed impact performance of a toughened T700/EH301 carbon/epoxy composite material. Three distinct hygrothermal conditions, namely Unaged, Aged, and Redried, are investigated. Experimental results reveal an apparent improvement in ballistic limit after hygrothermal aging (170.2 m/s, a notable increase of 11.2%), still maintaining an excellent performance even after redrying (158.5 m/s, an increase of 3.2%). Inspired by the micrographic analysis, a high-fidelity finite element model, referred to as HFEDCM (High Fidelity Explicit Discrete Crack Model), has been developed. This model not only explains the positive effects of hygrothermal aging on high-speed impact performance but also accurately captures the macroscopic damage morphology of composite laminates. It has been quantitatively and qualitatively demonstrated that the enhancement of the fracture toughness of the “fiber bundle” plays a critical role in improving the high-speed impact resistance of composites under hygrothermal aging.

Surface temperature field real-time reconstruction of hot forging die based on 1DCNN

Accurate reconstruction of the die surface temperature field during the hot forging process is crucial for ensuring optimal control of die temperature and high-quality metal forgings. In this study, we propose a novel method that combines hot forging process parameters with a deep neural network to reconstruct the temperature field on the die surface in real-time. Firstly, a high-fidelity finite element model of the hot forging process is created and validated with a forging experiment. Secondly, temperature observation points are initially selected based on the simulated temperature distribution of the forging die, and the correlation coefficient method is used to optimize the layout of observation points to improve reconstruction accuracy and efficiency. Subsequently, multiple high-quality simulations with varying process parameters are conducted, generating a comprehensive dataset. The 1DCNN algorithm is then presented to produce a mapping from the input to the surface temperature field. Finally, the effectiveness of the surface temperature field reconstruction method is verified. The results indicate that the overall performance of the proposed 1DCNN algorithm is better than the traditional regression algorithms. In terms of reconstruction accuracy, it achieves a prediction accuracy within an absolute error of 3 °C for the outer surface points and within a relative error of 5 % for the cavity region. Regarding reconstruction efficiency, it requires an approximate execution time of 0.064 s for predicting a new sample. This paper presents a method for real-time reconstruction of the die surface temperature field, leveraging hot forging process parameters and temperature data from observation points. This approach enables reliable online monitoring of transient temperature variations on the hot forging die surface, offering valuable support for operational control and maintenance.

Comparative analysis of rolling contact fatigue life in a wind turbine pitch bearing with different modeling approaches

The load calculation for wind turbines usually does not take pitch bearings into account, and the pitch-bearing internal load distribution is calculated in a decoupled process. In the current study, a coupled wind turbine load simulation considering the pitch bearings is proposed. Internal pitch-bearing load distribution and roller contact fatigue life are compared to a decoupled approach using a high-fidelity finite-element model. Internal pitch-bearing load distribution can be an essential variable in almost all major pitch-bearing failures. Furthermore, rolling contact fatigue is one of the major types of pitch-bearing failure that can represent the bearing service life. The results show that the flexibility of the surroundings has an effect on the contact forces. As a result, the finite element model-based life calculation of the NREL 5 MW turbine predicts a longer rolling contact fatigue life for the bearing than the multi-body system-based one. For the IWT7.5, sample load cases indicate that this behavior is different, leading to the conclusion that the life of the bearing is significantly affected by the stiffness of the bearing and its surrounding components.

Combined effect of random porosity and surface defect on fatigue lifetime of additively manufactured micro-sized Ti6Al4V components: An investigation based on numerical analysis and machine learning approach

Surface defects and internal porosities resulting from the additive manufacturing process contribute to a scatter in fatigue lifetime, increasing uncertainty in applications such as aerospace engineering. This study proposes a combined approach of machine learning and finite element modelling to explore the interaction between random porosity distribution and high surface roughness on the fatigue lifetime of micro-sized additively manufactured parts. To achieve this, a high-fidelity finite element model is constructed with randomly distributed porosities and surface defects. A fatigue lifetime estimation model, developed using a combination of continuum damage mechanics and the theory of critical distance, is then applied to predict lifetime using effective stress extracted from simulations. A total of 1200 simulations are conducted using these models to examine the interaction effects of subsurface porosity and surface defects on localized stress distribution and fatigue lifetime. Subsequently, two machine learning algorithms, Support Vector Regression and Kernel Ridge Regression, are employed to estimate fatigue lifetime, accounting for the effects of randomly distributed subsurface porosities and surface defects. Eighty percent of the developed finite element models are employed to construct an input dataset, incorporating geometrical features of pores, and to train the algorithms, while the remaining 20% are reserved for verification purposes. The observed results demonstrate that the combined effect of random porosity and surface defects increases fatigue scatter behaviour, leading to greater uncertainty in lifetime prediction. Furthermore, the machine learning algorithms exhibit promising prediction performance.

Effect of three-dimensional space on progressive collapse resistance of reinforced concrete frames under various column removal scenarios

In this study, high fidelity finite element (FE) model was adopted to investigate the progressive collapse resistance of reinforced concrete (RC) frame structures under various column removal scenarios. The validity of the numerical model was verified by comparing with the existing experimental results. Afterwards, to facilitate the research on possible simplified boundary conditions served for further various column removal scenarios, the effect of multi-span continuous beams was firstly evaluated and the simplified methods were compared. Then, the effect of various column removal scenarios addressing the three-dimensional effect on the progressive collapse resistance of RC frame structures with and without slab was systematically evaluated. The analysis results showed that the continuous span beams increased the internal compressive force of the beam section at compressive arch action (CAA) stage, which effectively improved the CAA strength of the substructure. For simple modeling, the use of horizontal constraints at the end of the beam could replace the continuous span beams. The presence of transverse beams and slab provided additional load transfer paths for collapse resistance. The compressive membrane action (CMA) and tensile membrane action (TMA) of the slab significantly improved the progressive collapse resistance of the substructure in the small and large deformation stages, respectively. The location of the removed column significantly affected the contribution ratio of transverse beams and slab to the progressive collapse resistance of the substructure, which was quantitatively evaluated for practical design.

Compression after impact (CAI) failure mechanisms and damage evolution in large composite laminates: High-fidelity simulation and experimental study

This study focuses on developing and validating a high-fidelity finite element model for predicting damage evolution and residual strength in fiber-reinforced composite panels. Impact and compression after impact (CAI) tests were conducted at both barely visible impact damage (BVID) and clearly visible impact damage (CVID) levels. The ASTM D7137 standard 100 mm × 150 mm CAI coupons were inadequate to cover the range of experimental studies required for model validation. Therefore, larger 254 mm × 304.8 mm laminates were investigated under two CAI testing conditions: one a scaled-up version of ASTM standard coupon, and the other with additional anti-buckling support plates to reduce unsupported areas to 127 mm × 177.8 mm. The model captured inter- and intra-laminar failure modes, including fiber breakage, splitting, kinking, pull-out, and crushing as well as matrix cracking, delamination, and their interactions. This was achieved by cohesive zone modeling technique and enhancement of the LaRC05 failure criteria through modeling the fiber damage evolution and utilizing an efficient search algorithm to determine the matrix fracture plane and fiber kink band angle. This study underscores the efficacy of the high-fidelity modeling approach in accurately predicting both impact damage and CAI strength in typical aircraft impact damage scenarios. Additionally, it provides insights into complex CAI failure mechanisms and energy release associated with various damage modes and highlights the effect of global buckling on the failure behavior and compressive strength of composite laminates. Furthermore, it shows that the proposed fixture with support plates is suitable for testing a broader range of impact scenarios without experiencing global buckling.

Risk assessment and prevention for typical railway bridge pier under rockfall impact

Railway bridges in mountainous regions are susceptible to damage and collapse due to rockfall impacts, posing a risk to train operations. This research specifically focuses on risk assessment of a typical railway bridge pier considering correlations between the collision parameters. Firstly, a comprehensive analysis of rockfall trajectories is conducted using the tool ROCFALL. Statistical features and correlation of impact parameters at the pier location are obtained and further characterized based on functional statistical models. Additionally, a high-fidelity finite element (FE) model of a two-span simply supported box girder with three piers is developed using LS-DYNA. Multiple structural dynamic responses including impact force, pier displacement, and internal energy of the directly impacted pier, are evaluated for both general rockfalls and potential boulders. The study examines the effects of individual impact parameters on the internal energy and displacement responses at the pier top, highlighting their significance in relation to load-bearing capacity and operational safety. Machine learning techniques are utilized to perform a comprehensive evaluation of risk states, wherein an Artificial Neural Network (ANN) model is employed for accurate prediction of risk indexes, and the K-means algorithm is applied to classify pier risk levels into minor, moderate, and severe. Furthermore, Monte Carlo (MC) simulations are employed to calculate the probabilities associated with different risk levels. A composite multilayer prevention measure is suggested to enhance the performance of a rockfall-prone bridge. It is proven that this measure can rapidly reduce associated vibrations and effectively mitigate structural damage resulting from the collision.

A Bayesian Multi-fidelity Neural Network to Predict Nonlinear Frequency Backbone Curves

Abstract The use of structural mechanics models during the design process often leads to the development of models of varying fidelity. Often low-fidelity models are efficient to simulate but lack accuracy, while the high-fidelity counterparts are accurate with less efficiency. This paper presents a multi-fidelity surrogate modeling approach that combines the accuracy of a high-fidelity finite element model with the efficiency of a low-fidelity model to train an even faster surrogate model that parameterizes the design space of interest. The objective of these models is to predict the nonlinear frequency backbone curves of the Tribomechadynamics Research Challenge benchmark structure which exhibits simultaneous nonlinearities from frictional contact and geometric nonlinearity. The surrogate model consists of an ensemble of neural networks that learn the mapping between low and high-fidelity data through nonlinear transformations. Bayesian neural networks are used to assess the surrogate model's uncertainty. Once trained, the multi-fidelity neural network is used to perform sensitivity analysis to assess the influence of the design parameters on the predicted backbone curves. Additionally, Bayesian calibration is performed to update the input parameter distributions to correlate the model parameters to the collection of experimentally measured backbone curves.

Error approximation and bias correction in dynamic problems using a recurrent neural network/finite element hybrid model

This work proposes a hybrid modeling framework based on recurrent neural networks (RNNs) and the finite element (FE) method to approximate model discrepancies in time-dependent, multi-fidelity problems, and use the trained hybrid models to perform bias correction of the low-fidelity models. The hybrid model uses FE basis functions as a spatial basis and RNNs for the approximation of the time dependencies of the FE basis' degrees of freedom. The training data sets consist of sparse, non-uniformly sampled snapshots of the discrepancy function, pre-computed from trajectory data of low- and high-fidelity dynamic FE models. To account for data sparsity and prevent overfitting, data upsampling and local weighting factors are employed, to instigate a trade-off between physically conforming model behavior and neural network regression. The proposed hybrid modeling methodology is showcased in three highly non-trivial engineering test-cases, all featuring transient FE models, namely, heat diffusion out of a heat sink, eddy-currents in a quadrupole magnet, and sound wave propagation in a cavity. The results show that the proposed hybrid model is capable of approximating model discrepancies to a high degree of accuracy and accordingly correct low-fidelity models.

Experimental tests and numerical simulations of miniature buckling-restrained braces using iron-shape memory alloy bar

Miniature buckling-restrained braces (BRBs) have a concise configuration and clear working mechanism. Owing to stable and full hysteresis, miniature BRBs are emerging as favorable energy-dissipating fuse elements in seismic applications. This paper proposes using iron-shape memory alloy (FeSMA) in lieu of conventional steel as the yielding core and focuses on the cyclic behavior. Four specimens, corresponding to four loading protocols, were fabricated and tested. According to experimental data, the miniature FeSMA BRBs exhibited full hysteresis, which is characterized by large damping, noticeable isotropic and kinematic hardening behavior and appreciable post-yield stiffness ratio. The cumulative plastic deformations meet the requirement by ANSI/AISC 341-16. Further, high fidelity finite element models, which explicitly considered damage and fracture rules, were established for numerical simulations. Good agreement can be found between the experimental data and numerical simulations, in terms of failure modes, fracture locations and global hysteresis.

Collapse-resistant mechanisms induced by various perimeter column damage scenarios in RC flat plate structures

This paper aims at assessing the impact of removing various numbers of perimeter columns on the integrity of RC flat plate substructures. Compared to interior columns, those at the corner and edge positions on the building perimeter are more likely to damage by accidental and abnormal loads, which in turn may cause progressive collapse of the entire structure. Given the uncertainty of the loading events, the likelihood of multiple perimeter columns being simultaneously destroyed is also high. However, existing research has primarily focused on the single column removal cases. To address this research limitation, a 3D high-fidelity finite element model of a flat plate substructure with a missing corner column (named Model C) was developed. The accuracy of the model was validated based on our previous experiment of the same substructure. The validated model was subsequently modified to cover two and multiple perimeter column removal scenarios. The two new models are namely, Model C_1e in which a corner column and an adjacent edge column were damaged and Model C_2e, with a corner column and two adjacent edge columns being removed. The simulation results reveal that flexural action was the primary load resisting mechanism in all three models. The load carrying capacity with three column removals was reduced by 75.6% compared to the removal of a single corner column. Full-depth cracks were found at the cut-off position of the top slab rebars for the columns adjacent to the removed one(s). In Model C and Model C_2e, cracks exhibited a diagonal axisymmetric failure pattern of the corner of the slab, with equal load transfer to two adjacent columns. In contrast, Model C_1e exhibited a diagonal non-axisymmetric failure pattern of the slab corner, with more loads being transferred to the adjacent column along the shorter bay. The energy-based dynamic response analyses were also performed to obtain the Dynamic Increase Factor (DIF) for the substructure models under different column removal scenarios. Based on the nonlinear regression analysis, the DIFs for the substructures with one and two column removals were found to be ranged between 1.22 and 1.51, and 1.28–1.68, respectively.

Cyclic loading tests and numerical simulations of an assembled X-shaped flexural steel damper

This paper proposes an assembled X-shaped flexural steel damper. Welding is waived for the novel damper. As the kernel element of the damper, the X-shaped flexural steel plates have the advantages of ease of installation and replacement if necessary. In the analytical part, the effective height of the flexural steel plate was suggested, and then, the yield displacement, yield load and initial stiffness of the damper were derived. The proof-of-concept cyclic loading test was conducted using one reduced-scale specimen within a four-bar linkage experimental setup. The hysteretic performance indexes, including strength, stiffness and energy dissipation, were quantified. The high-fidelity finite element (FE) models were established by ABAQUS. The parametric study was further conducted based on the verified FE model. The considered parameters were the height and thickness of the steel plates. The testing results indicate that the proposed damper has plump hysteretic curve with good energy dissipation capacity and excellent ductility, although slight slip behavior can be found due to the gap caused by the assembling process. The results obtained from analytical method and numerical model are in good agreement with the experimental data. The results of parametric analysis show that the numerical results well agree with analytical predictions, indicating the numerical model can reflect the mechanical behavior of AXSFD. A benchmark frame building was selected to demonstrate the seismic control efficacy of the damper. The nonlinear time history analysis results indicate that the damper reduced both peak and residual interstory drift ratios for the protected structure.

Superelastic Shape Memory Alloy Honeycomb Damper

The relative displacements between the girders and piers of isolated bridges during intense earthquakes are usually so large that traditional restrainers cannot accommodate the resulting deformation. A novel superelastic shape memory alloy (SMA) honeycomb damper (SHD) is proposed as a means to combine the large strain capacity of SMA and the geometrical nonlinear deformation of honeycomb structures. As a result, the large deformation capacity of the novel damper satisfies the requirements for bridge restrainers. The proposed device consists of a superelastic shape memory alloy (SMA) honeycomb structure, which enables a self-centering capability, along with steel plates that serve to prevent the buckling of the SMA honeycomb. An examination of the SHD was undertaken initially from theoretical perspectives. A multi-cell SHD specimen was subsequently manufactured and evaluated. Following this, numerical simulation analyses of the SHDs using a three-dimensional high-fidelity finite element model were employed to examine the experimental results. In the end, a technique for improving the SHD was suggested. The results indicate that the SHD is able to demonstrate superior self-centering capabilities and stable hysteretic responses when subjected to earthquakes.

High-fidelity simulations of low-velocity impact induced matrix cracking and dynamic delamination progression in CFRP beams

A high-fidelity finite element model is constructed to simulate the low-velocity line-impact experiments conducted by Bozkurt and Coker on [05/903]s CFRP beams. The simulations utilize a user-implemented three-dimensional continuum damage model with the LaRC05 criterion for matrix cracking, and a built-in cohesive zone model for delamination in ABAQUS/Explicit. The significant influence of boundary supports on the global impact response leads to proposing a heuristic boundary conditions approach using spring elements that replicate the experiment boundaries. Results then demonstrated an excellent agreement with the experiments in terms of the global impact response, the strain field, and damage pattern and sequence. Simulations reveal intersonic delamination with crack tip speeds around ∼5000 m/s, while the experimental crack speeds were measured as sub-Rayleigh speeds reaching ∼1000 m/s. When a crack tip definition based on the crack tip opening is introduced in the simulations, crack tip speeds in the range of 490–1500 m/s are measured which are within the range of experimental speeds, suggesting that the sliding mode might be physically hidden in the experiments. The potential use of delamination crack tip speeds as a benchmark for refining numerical simulations is demonstrated by increasing the effective interface toughness leading to a decrease in crack tip speeds with no noticeable effect on the global responses.

An efficient framework for structural seismic collapse capacity assessment based on an equivalent SDOF system

Assessing the structural collapse capacity efficiently and accurately is a key issue in earthquake engineering. Here, an efficient framework for structural seismic collapse capacity assessment based on an accurate equivalent single-degree-of-freedom (ESDOF) system is proposed. The corresponding ESDOF system is calibrated by matching the cyclic static pushover (SPO) curves of a more complex multi-degree-of-freedom (MDOF) system (e.g., high-fidelity finite element models) through a meta-heuristic optimization method. In this way, both the backbone curve and hysteretic characteristics (the stiffness and strength deterioration, and pinching behavior) of the complex MDOF system are considered in the corresponding ESDOF system. Then, incremental dynamic analysis (IDA) is carried out to assess the structural seismic collapse capacity using the ESDOF system instead of the corresponding MDOF system to improve the computational efficiency. The efficiency and accuracy of the proposed framework have been validated through three case studies, including a bare reinforced concrete (RC) frame, a steel frame, and an infilled wall RC frame. These results affirm that the proposed framework can accurately and efficiently assess the collapse capacity of low-rise bare and infilled RC frame structures, as well as steel frame structures.

Collapse-resistant mechanism of RC beam-slab substructures using kinked steel plates under two-adjacent-edge-columns removal scenario

To improve the progressive collapse resistance of reinforced concrete (RC) frame structures without basically affecting their seismic performance, the authors have proposed a novel method of adding locally debonded kinked steel plates (KPs) at the beam-ends of the RC frames. This method has been verified through quasi-static tests on the RC beam-column assemblies. Considering the presence of slabs, this paper investigated the collapse-resistant mechanism and resistance improvement for the RC beam-slab substructures using locally debonded KPs. First, quasi-static tests were conducted on two RC beam-slab substructures under the scenario of removing two-adjacent-edge-columns. One substructure was configurated with the KPs and the other without the KPs for comparison. Test results demonstrated that, by adding the KPs, the ultimate resistance and deformability of the beam-slab substructure improved by 76% and 40%, respectively. Second, the high-fidelity finite element (FE) models were created to analyze the load transfer path of the substructure and the resistance contribution at the stage of catenary mechanism. Combining the test and FE analysis results, the collapse-resistant mechanism was revealed, i.e., the KPs could fully mobilize and develop the catenary action in the edge-beams and the tensile membrane action in the slabs along the edge direction. Finally, an analytical model was developed for predicting the ultimate resistance of the RC beam-slab substructures with or without the KPs under the scenario of removing two-adjacent-edge-columns.

Machine learning modeling of lung mechanics: Assessing the variability and propagation of uncertainty in respiratory-system compliance and airway resistance

Background and ObjectiveTraditional assessment of patient response in mechanical ventilation relies on respiratory-system compliance and airway resistance. Clinical evidence has shown high variability in these parameters, highlighting the difficulty of predicting them before the start of ventilation therapy. This motivates the creation of computational models that can connect structural and tissue features with lung mechanics. In this work, we leverage machine learning (ML) techniques to construct predictive lung function models informed by non-linear finite element simulations, and use them to investigate the propagation of uncertainty in the lung mechanical response. MethodsWe revisit a continuum poromechanical formulation of the lungs suitable for determining patient response. Based on this framework, we create high-fidelity finite element models of human lungs from medical images. We also develop a low-fidelity model based on an idealized sphere geometry. We then use these models to train and validate three ML architectures: single-fidelity and multi-fidelity Gaussian process regression, and artificial neural networks. We use the best predictive ML model to further study the sensitivity of lung response to variations in tissue structural parameters and boundary conditions via sensitivity analysis and forward uncertainty quantification. Codes are available for download at https://github.com/comp-medicine-uc/ML-lung-mechanics-UQ ResultsThe low-fidelity model delivers a lung response very close to that predicted by high-fidelity simulations and at a fraction of the computational time. Regarding the trained ML models, the multi-fidelity GP model consistently delivers better accuracy than the single-fidelity GP and neural network models in estimating respiratory-system compliance and resistance (R2∼0.99). In terms of computational efficiency, our ML model delivers a massive speed-up of ∼970,000× with respect to high-fidelity simulations. Regarding lung function, we observed an almost matched and non-linear behavior between specific structural parameters and chest wall stiffness with compliance. Also, we observed a strong modulation of airways resistance with tissue permeability. ConclusionsOur findings unveil the relevance of specific lung tissue parameters and boundary conditions in the respiratory-system response. Furthermore, we highlight the advantages of adopting a multi-fidelity ML approach that combines data from different levels to yield accurate and efficient estimates of clinical mechanical markers. We envision that the methods presented here can open the way to the development of predictive ML models of the lung response that can inform clinical decisions.