Search

Fracture of materials with rate-dependent mechanical behaviour, e.g. polymers, is a highly complex process. For an adequate modelling, the coupling between rate-dependent stiffness, dissipative mechanisms present in the bulk material and crack driving force has to be accounted for in an appropriate manner. In addition, the resistance against crack propagation can depend on rate of deformation. In this contribution, an energetic phase-field model of rate-dependent fracture at finite deformation is presented. For the deformation of the bulk material, a formulation of finite viscoelasticity is adopted with strain energy densities of Ogden type assumed. The unified formulation allows to study different expressions for the fracture driving force. Furthermore, a possibly rate-dependent toughness is incorporated. The model is calibrated using experimental results from the literature for an elastomer and predictions are qualitatively and quantitatively validated against experimental data. Predictive capabilities of the model are studied for monotonic loads as well as creep fracture. Symmetrical and asymmetrical crack patterns are discussed and the influence of a dissipative fracture driving force contribution is analysed. It is shown that, different from ductile fracture of metals, such a driving force is not required for an adequate simulation of experimentally observable crack paths and is not favourable for the description of failure in viscoelastic rubbery polymers. Furthermore, the influence of a rate-dependent toughness is discussed by means of a numerical study. From a phenomenological point of view, it is demonstrated that rate-dependency of resistance against crack propagation can be an essential ingredient for the model when specific effects such as rate-dependent brittle-to-ductile transitions shall be described. [ABSTRACT FROM AUTHOR]

Understanding tissue rheology is critical to accurately model the human heart. While the elastic properties of cardiac tissue have been extensively studied, its viscous properties remain an issue of ongoing debate. Here we adopt a viscoelastic version of the classical Holzapfel Ogden model to study the viscous timescales of human cardiac tissue. We perform a series of simulations and explore stress–relaxation curves, pressure–volume loops, strain profiles, and ventricular wall strains for varying viscosity parameters. We show that the time window for model calibration strongly influences the parameter identification. Using a four-chamber human heart model, we observe that, during the physiologically relevant time scales of the cardiac cycle, viscous relaxation has a negligible effect on the overall behavior of the heart. While viscosity could have important consequences in pathological conditions with compromised contraction or relaxation properties, we conclude that, for simulations within the physiological range of a human heart beat, we can reasonably approximate the human heart as hyperelastic. [ABSTRACT FROM AUTHOR]

We present an approach for the data-driven modeling of nonlinear viscoelastic materials at small strains which is based on physics-augmented neural networks (NNs) and requires only stress and strain paths for training. The model is built on the concept of generalized standard materials and is therefore thermodynamically consistent by construction. It consists of a free energy and a dissipation potential, which can be either expressed by the components of their tensor arguments or by a suitable set of invariants. The two potentials are described by fully/partially input convex neural networks. For training of the NN model by paths of stress and strain, an efficient and flexible training method based on a long short-term memory cell is developed to automatically generate the internal variable(s) during the training process. The proposed method is benchmarked and thoroughly compared with existing approaches. Different databases with either ideal or noisy stress data are generated for training by using a conventional nonlinear viscoelastic reference model. The coordinate-based and the invariant-based formulation are compared and the advantages of the latter are demonstrated. Afterwards, the invariant-based model is calibrated by applying the three training methods using ideal or noisy stress data. All methods yield good results, but differ in computation time and usability for large data sets. The presented training method based on a recurrent cell turns out to be particularly robust and widely applicable. We show that the presented model together with the recurrent cell for training yield complete and accurate 3D constitutive models even for sparse bi- or uniaxial training data. [ABSTRACT FROM AUTHOR]

In this work, the geometry-constraint-based non-ordinary state-based peridynamic (GC-NOSBPD) model is further generalized to study the mechanical responses and rate-dependent fracture behaviors of viscoelastic materials. The viscoelasticity is generated by directly incorporating the classical viscoelastic constitutive relation in the form of the Generalized Maxwell Model, which is expressed as Prony-series type expansion, into the GC-NOSBPD framework. A strain-based scalar damage variable only driven by elastic energy, whose evolution follows a rate-independent law, is proposed to describe the progressive growth of the mode-I crack. Combining the rate-independent damage variable and viscoelasticity results in a rate-dependent damage model in the hypothesis that the fracture of viscoelastic material is driven by both elastic and viscous components of the energy. The fidelity of this model is established in the absence of damage by comparison with the benchmark solution for the viscoelastic response of a 3D bar subjected to various load histories. Further, the proposed model is employed to simulate two quasi-static fracture tests under various loading rates: a three-point bend fracture test for concrete and a double-cantilever-beam delamination test for rubber interface. These numerical results agree well with experimental tests, showing that the proposed model can effectively capture rate-dependent crack propagation. [ABSTRACT FROM AUTHOR]

Within a 3D concrete printing process, the fresh concrete is aging due to hydration. One of the consequences from the purely mechanical point of view is that its constitutive relation must be defined in rate form. This restriction is taken into account in this contribution and, besides on the incremental elasticity, we moreover introduce the relaxation of the internal stresses in order to describe the creep at early age. On another hand, due to the soft nature of the material, the finite strain range is herein a priori assumed. Eventual structural instabilities during the printing process can therefore be predicted as well. On another hand, with regards to the incremental formulation of the boundary value problem, the kinematics must be adapted as well. We use for this the multiplicative decomposition of the actual deformation gradient into its known part at an earlier time and the relative deformation gradient with respect to the configuration at that time. Within a Lagrangian formulation, the incremental constitutive relations and evolution equations can then be ideally defined on the above mentioned intermediate configuration prior to be transported back to the reference configuration. In particular, the early age creep is here described through an internal variable approach the evolution of which is motivated by the generalized Maxwell model. In this work, this latter is adapted for incremental viscoelasticity. Model examples are proposed and the numerical efficiency of the proposed framework is illustrated through a set of representative simulations. [ABSTRACT FROM AUTHOR]

In this work, a consistent viscoplasticity formulation is derived from thermodynamical principles and employing the concept of continuum elastic corrector rate. The proposed model is developed based on the principle of maximum viscoplastic dissipation for determining the flow direction. The model uses both the equivalent viscoplastic strain and its rate as state variables. Power balance and energy balance give, respectively, separate evolution equations for the equivalent viscoplastic strain rate and the viscoplastic strain, the former written in terms of inviscid rates. Several key points distinguish our formulation from other proposals. First, the viscoplastic strain rate (instead of a yield function) consistently distinguishes conservative from dissipative behaviours during reverse loading; and the discrete implicit integration algorithm is an immediate discretization of the continuum theory based on the mentioned principles considered as separate equations. Second, the inviscid solution is recovered in a well-conditioned manner by simply setting the viscosity to zero. Indeed, inviscid plasticity, viscoelasticity and viscoplasticity are particular cases of our formulation and integration algorithm, and are recovered just by setting the corresponding parameters to zero (viscosity or yield stress). Third, the linear viscoplasticity solution is obtained in an exact manner for proportional loading cases, independently of the time step employed. Four, general nonlinear models (Perzyna, Norton, etc) may be immediately incorporated as particular cases both in the theory and the computational implementation. [ABSTRACT FROM AUTHOR]

Understanding how axons fail is critical to preventing brain injury. From stretch experiments, we know how axons respond to forces on the time scales of milliseconds and days. Yet, there is no mechanical model that explains the behavior of the axon at both short and long time scales. Here we propose a constitutive model to study the limits of stretch-mediated axonal disconnection at different time scales. Our model combines viscoelasticity using a neo-Hookean standard linear solid and growth using stress-mediated accelerated elongation. By limiting peak and average membrane tensions, our model predicts critical elongations and elongation rates. Interestingly, the critical elongation rate is not constant, but increases after an acclimation period. Combining viscoelasticity and growth is essential to simulate axonal disconnection in stretch-mediated growth at both short and long time scales. Our model can help optimize axonal stretch experiments and provides insight into the interacting time scales within the axon. [ABSTRACT FROM AUTHOR]

The phase-field model has been employed for fracture analysis over the last decade. One of the main advantages is that the fracture evolution does not depend on any explicit criterion. Good agreement is obtained by comparing the results of simulations and experiments. Most elastomers exhibit both elastic and viscous behavior simultaneously, which yields rate-dependent properties for both the mechanical and fracture responses. In this contribution, a viscoelastic rheological model based on Reese and Govindjee (Int J Solids Struct 35:3455–3482, 1998) is coupled to phase-field modeling to investigate rate-dependent fracture within elastomers. The elastic strain energy potential supposed to evolve fracture is provided by both the equilibrium and non-equilibrium branches. The fracture mechanism is characterized by a volumetric–isochoric split, which specifies a varying driving force in the case of tensile or compressive deformation. Representative numerical examples are studied and related findings and potential perspectives are summarized to close the paper. [ABSTRACT FROM AUTHOR]

Deformation of soft polymeric materials often involves complex nonlinear or transient mechanical behaviors. This is due to the dynamic behaviors of polymer chains at the molecular level within the polymer network. In this paper, we present a computational formulation to describe the transient behavior (e.g., viscoelasticity) of soft polymer networks with dynamic bonds undergoing large to extreme deformation. This formulation is based on an Eulerian description of kinematics and a theoretical framework that directly connects the molecular-level kinetics of dynamic bonds to the macroscopic mechanical behavior of the material. An extended finite element method is used to discretize the field variables and the governing equations in an axisymmetric domain. In addition to validating the framework, this model is used to study how the chain dynamics affect the macroscopic response of material as they undergo a combination of flow and elasticity. The problems of cavitation rheology and polymer indentation under extreme deformation are investigated in this context. [ABSTRACT FROM AUTHOR]

This paper deals with the computational homogenization and numerical model reduction of deformation driven pressure diffusion in fractured porous rock. Exposed to seismic waves, the heterogeneity of the material leads to local fluid pressure gradients which are equilibrated via pressure diffusion. However, a macroscopic observer is not able to measure the diffusion process directly but senses the intrinsic attenuation of an apparently monophasic viscoelastic solid. The aim of this paper is to establish a reliable, yet numerically efficient, computational homogenization method to identify the viscoelastic properties of the macroscopic substitute model. Inspired by the Nonuniform Transformation Field Analysis, we incorporate a Numerical Model Reduction procedure. The proposed method is validated for several scenarios ranging from pressure diffusion in an unfractured poroelastic matrix, via localized pressure diffusion in interconnected fractures embedded in an impermeable matrix, to the fully coupled pressure diffusion both in fractures and the embedding poroelastic matrix. [ABSTRACT FROM AUTHOR]

In order to reduce the transfer of sound and vibrations in structures such as timber buildings, thin elastomer layers can be embedded between their components. The influence of these elastomers on the response of the structures in the low frequency range can be determined accurately by using conforming hexahedral finite elements. Three-dimensional mesh generation, however, is yet a non-trivial task and mesh refinements which may be necessary at the junctions can cause a high computational effort. One remedy is to mesh the components independently from each other and to couple them using the mortar method. Further, the hexahedral mesh for the thin elastomer layer itself can be avoided by integrating its elastic behavior into the mortar formulation. The present paper extends this mortar formulation to take damping into account such that frequency response analyses can be performed more accurately. Finally, the proposed method is verified by numerical examples. [ABSTRACT FROM AUTHOR]

Working mechanisms of the cardiac defibrillation are still in debate due to the limited experimental facilities and one-third of patients even do not respond to cardiac resynchronization therapy. With an aim to develop a milestone towards reaching the unrevealed mechanisms of the defibrillation phenomenon, we propose a bidomain based finite element formulation of cardiac electromechanics by taking into account the viscous effects that are disregarded by many researchers. To do so, the material is deemed as an electro-visco-active material and described by the modified Hill model (Cansız et al. in Comput Methods Appl Mech Eng 315:434-466, 2017). On the numerical side, we utilize a staggered solution method, where the elliptic and parabolic part of the bidomain equations and the mechanical field are solved sequentially. The comparative simulations designate that the viscoelastic and elastic formulations lead to remarkably different outcomes upon an externally applied electric field to the myocardial tissue. Besides, the achieved framework requires significantly less computational time and memory compared to monolithic schemes without loss of stability for the presented examples. [ABSTRACT FROM AUTHOR]

A structure-preserving time integration scheme is a modified standard time stepping scheme, which fulfills the balance laws of a dynamical problem also in a discrete setting. Hence, such a scheme represents a space-time discretization of a continuum with all of its central properties. This paper deals with a special second-order accurate structure-preserving time integration of thermo-mechanical couplings in a moving continuum, satisfying all discrete thermodynamical balance laws in contrast to the underlying second-order accurate midpoint rule. More accurately, we consider isotropic heat conduction and the Gough-Joule effect in moving continua. The considered time evolution is described by five differential equations with respect to five independent variables, including the entropy field of the continuum. The bodies in the numerical examples are subjected to Dirichlet- or Neumann boundary conditions in the displacement as well as the temperature field. Although this structure-preserving time integrator is based on a time discrete spatially weak finite element formulation, it fulfills the same balance laws as the underlying five differential equations. Namely, in addition to the balances of entropy, linear momentum and angular momentum, also the balances of total energy and Lyapunov’s function are fulfilled in the discrete case. [ABSTRACT FROM AUTHOR]

For short fiber reinforced plastic parts the local fiber orientation has a strong influence on the mechanical properties. To enable multiscale computations using surrogate models we advocate a two-step identification strategy. Firstly, for a number of sample orientations an effective model is derived by numerical methods available in the literature. Secondly, to cover a general orientation state, these effective models are interpolated. In this article we develop a novel and effective strategy to carry out this interpolation. Firstly, taking into account symmetry arguments, we reduce the fiber orientation phase space to a triangle in R2. For an associated triangulation of this triangle we furnish each node with an surrogate model. Then, we use linear interpolation on the fiber orientation triangle to equip each fiber orientation state with an effective stress. The proposed approach is quite general, and works for any physically nonlinear constitutive law on the micro-scale, as long as surrogate models for single fiber orientation states can be extracted. To demonstrate the capabilities of our scheme we study the viscoelastic creep behavior of short glass fiber reinforced PA66, and use Schapery’s collocation method together with FFT-based computational homogenization to derive single orientation state effective models. We discuss the efficient implementation of our method, and present results of a component scale computation on a benchmark component by using ABAQUS ®. [ABSTRACT FROM AUTHOR]

The present paper is the second part of a twofold work, whose first part is reported in Artioli et al. (Comput Mech, 2017. doi:), concerning a newly developed Virtual element method (VEM) for 2D continuum problems. The first part of the work proposed a study for linear elastic problem. The aim of this part is to explore the features of the VEM formulation when material nonlinearity is considered, showing that the accuracy and easiness of implementation discovered in the analysis inherent to the first part of the work are still retained. Three different nonlinear constitutive laws are considered in the VEM formulation. In particular, the generalized viscoelastic model, the classical Mises plasticity with isotropic/kinematic hardening and a shape memory alloy constitutive law are implemented. The versatility with respect to all the considered nonlinear material constitutive laws is demonstrated through several numerical examples, also remarking that the proposed 2D VEM formulation can be straightforwardly implemented as in a standard nonlinear structural finite element method framework. [ABSTRACT FROM AUTHOR]

The paper is devoted to non-linear vibrations of plates, made of the Zener viscoelastic material modelled with the Caputo fractional derivative, and in particular to their response to harmonic excitation. The plate geometric non-linearity is of the von Kármán type. In the formulation shear effects and rotary inertia are considered, too. The problem is solved in the frequency domain. A one-harmonic form of the solution for plate displacements corresponding to the plate formulation is assumed. The amplitude equation is obtained from the time averaged principle of virtual work. The time averaging precedes the use of the harmonic balance method. In the space discretization the finite element method is used involving 8-noded rectangular plate elements with selective-reduced integration. Several numerical examples are analyzed and the response curves are found using a path-following method. The purpose of these analyses is to identify material features of the adopted model of viscoelasticity with the fractional derivative. [ABSTRACT FROM AUTHOR]

Polymer nanocomposites have been envisioned as advanced materials for improving the mechanical performance of neat polymers used in aerospace, petrochemical, environment and energy industries. With the filler size approaching the nanoscale, composite materials tend to demonstrate remarkable thermomechanical properties, even with addition of a small amount of fillers. These observations confront the classical composite theories and are usually attributed to the high surface-area-to-volume-ratio of the fillers, which can introduce strong nanoscale interfacial effect and relevant long-range perturbation on polymer chain dynamics. Despite decades of research aimed at understanding interfacial effect and improving the mechanical performance of composite materials, it is not currently possible to accurately predict the mechanical properties of polymer nanocomposites directly from their molecular constituents. To overcome this challenge, different theoretical, experimental and computational schemes will be used to uncover the key physical mechanisms at the relevant spatial and temporal scales for predicting and tuning constitutive behaviors in silico, thereby establishing a bottom-up virtual design principle to achieve unprecedented mechanical performance of nanocomposites. A modular-based multiscale modeling approach for viscoelasticity of polymer nanocomposites has been proposed and discussed in this study, including four modules: (A) neat polymer toolbox; (B) interphase toolbox; (C) microstructural toolbox and (D) homogenization toolbox. Integrating these modules together, macroscopic viscoelasticity of polymer nanocomposites could be directly predicted from their molecular constituents. This will maximize the computational ability to design novel polymer composites with advanced performance. More importantly, elucidating the viscoelasticity of polymer nanocomposites through fundamental studies is a critical step to generate an integrated computational material engineering principle for discovering and manufacturing new composites with transformative impact on aerospace, automobile, petrochemical industries. [ABSTRACT FROM AUTHOR]

This paper discusses the finite-deformation viscoelastic modeling for passive myocardium tissue. The formulations established can also be applied to model other fiber-reinforced soft tissue. Based on the morphological structure of the myocardium, a specific free-energy function is constructed to reflect its orthotropicity. After deriving the stress-strain relationships in the simple shear deformation, a genetic algorithm is used to optimally estimate the material parameters of the myocardial constitutive equation. The results show that the proposed myocardial model can well fit the shear experimental data. To validate the viscoelastic model, it is used to predict the creep and the dynamic responses of a cylindrical model of the left ventricle. Upon comparing the results calculated by the proven myocardial elastic model with those by the viscoelastic model, the merits of the latter are discussed. [ABSTRACT FROM AUTHOR]

In order to gain a deeper knowledge of the interactions in the coupled tire-pavement-system, e.g. for the future design of durable pavement structures, the paper presents recent results of research in the field of theoretical-numerical asphalt pavement modeling at material and structural level, whereby the focus is on a realistic and numerically efficient computation of pavements under rolling tire load by using the finite element method based on an Arbitrary Lagrangian Eulerian (ALE) formulation. Inelastic material descriptions are included into the ALE frame efficiently by a recently developed unsplit history update procedure. New is also the implementation of a viscoelastic cohesive zone model into the ALE pavement formulation to describe the interaction of the single pavement layers. The viscoelastic cohesive zone model is further extended to account for the normal pressure dependent shear behavior of the bonding layer. Another novelty is that thermo-mechanical effects are taken into account by a coupling of the mechanical ALE pavement computation to a transient thermal computation of the pavement cross-section to obtain the varying temperature distributions of the pavement due to climatic impact. Then, each ALE pavement simulation considers the temperature dependent asphalt material model that includes elastic, viscous and plastic behavior at finite strains and the temperature dependent viscoelastic cohesive zone formulation. The temperature dependent material parameters of the asphalt layers and the interfacial layers are fitted to experimental data. Results of coupled tire-pavement computations are presented to demonstrate potential fields of application. [ABSTRACT FROM AUTHOR]

There are many recent studies concerning reduced-order computational methods, especially reductions by projection on a small-sized basis. But it is difficult to control the quality of the solutions if the basis is fixed once and for all. This is why we attempt to define efficient and low-cost strategies for correction and updating of the basis. These correction steps re-use previously computed quantities such as: vectors and triangulated matrices. The proposed algorithms use alternately full and reduced-size steps, allowing a strong reduction in the number of full-size tangent matrices. Two classes of applications are discussed. First, we consider an algorithm for determining Hopf bifurcation points in 2D Navier-Stokes equations, but which requires time-consuming preliminary frequency-dependent calculations. New reduction-correction procedures are applied to reduce these preliminary computations. The second application concerns the response curves of viscoelastic structures. A key point is the definition of the reduced basis. Vector Taylor series are computed within the asymptotic numerical method and the relevance of this set of vectors is analysed. [ABSTRACT FROM AUTHOR]

In this contribution, the convergence behaviour of simulations with nonlinear viscoelastic models for rubber-like materials using fractional derivatives is investigated. Based on the complex step derivative approximation, numerical approximation schemes for tangent operators on the local and global algorithmic level are analysed. Material models including fractional derivatives usually exhibit numerical difficulties, since the entire stress history of the material has to be considered for the current stress state. Non-classical methods can help to reduce the numerical effort, but the convergence behaviour is not perfect. In this paper, a classical and a non-classical fractional element based viscoelastic material formulation are analysed. The local and the global convergence rate using analytically and numerically derived tangents are investigated and compared to the convergence behaviour of a standard nonlinear viscoelastic model. Numerical examples on rubber materials are exploited, showing the performance of the proposed methods. It can be proven that the outlined numerical differentiation schemes improve the convergence rate, as well as reduce the computation time for the fractional element based material models. [ABSTRACT FROM AUTHOR]

In this paper a purely phenomenological formulation and finite element numerical implementation for quasi-incompressible transversely isotropic and orthotropic materials is presented. The stored energy is composed of distinct anisotropic equilibrated and non-equilibrated parts. The nonequilibrated strains are obtained from the multiplicative decomposition of the deformation gradient. The procedure can be considered as an extension of the Reese and Govindjee framework to anisotropic materials and reduces to such formulation for isotropic materials. The stress-point algorithmic implementation is based on an elastic-predictor viscous-corrector algorithm similar to that employed in plasticity. The consistent tangent moduli for the general anisotropic case are also derived. Numerical examples explain the procedure to obtain the material parameters, show the quadratic convergence of the algorithm and usefulness in multiaxial loading. One example also highlights the importance of prescribing a complete set of stress-strain curves in orthotropic materials. [ABSTRACT FROM AUTHOR]

This paper deals with the development of the linear vibration of a general viscoelastic structure, with a local wall acoustic impedance, containing an inviscid compressible liquid (but with an additional volume dissipative term), with surface tension (capillarity) and sloshing effects, and neglecting the effects of internal gravity waves and the elastogravity operator. The sloshing problems of incompressible liquids with capillarity effects in elastic structures exhibit a major difficulty induced by the boundary contact conditions on the triple line because the capillarity forces are forces per unit length while the elastic forces are forces per unit surface. The proposed framework has the following novel features: (i) introducing a new appropriate boundary condition for the contact angle condition compatible with a deformable structure considered here as viscoelastic, (ii) considering a compressible liquid while incompressibility hypothesis is generally used for FSI problems including capillarity phenomena, and (iii) constructing a reduced-order model for the computational coupled problem. [ABSTRACT FROM AUTHOR]

This contribution aims at achieving two important goals: First, it outlines a numerical inverse homogenization strategy able to recover material parameters of the microstructure by using results of macroscopic tests. Second, it considers parameter identification for viscoelastic heterogeneous materials, which is a step providing the basis for the further extension toward the general treatment of dissipative processes. The approach proposed couples the Levenberg-Marquardt method with the multiscale finite element method. In this combination, the former is a gradient-based optimization strategy used to minimize a merit function while the latter is a numerical homogenization technique needed to solve the direct problem. The specific example studied in the paper deals with the investigation of a composite consisting of a viscoelastic curing polymer and a nonlinear elastic material. It proposes a three-step procedure for the evaluation of its material parameters and discusses the accuracy and the uniqueness of the solution. [ABSTRACT FROM AUTHOR]

This paper is concerned with the numerical solution of the evolution equations of thermomechanical systems, in such a way that the scheme itself satisfies the laws of thermodynamics. Within this framework, we present a novel integration scheme for the dynamics of viscoelastic continuum bodies in isothermal conditions. This method intrinsically satisfies the laws of thermodynamics arising from the continuum, as well as the possible additional symmetries. The resulting solutions are physically accurate since they preserve the fundamental physical properties of the model. Furthermore, the method gives an excellent performance with respect to robustness and stability. Proof for these claims as well as numerical examples that illustrate the performance of the novel scheme are provided. [ABSTRACT FROM AUTHOR]

Recurrent neural networks (RNNs) have demonstrated very impressive performances in learning sequential data, such as in language translation and music generation. Here, we show that the intrinsic computational aspect of RNNs is very similar to that of classical stress update algorithms in modeling history-dependent materials with an emphasis on viscoelasticity. Several numerical examples are designed, including 1-dimensional and 3-dimensional cases, which testify the ability of RNN model to compute the viscoelastic response when predicting on unseen test data. Additionally, it is found that the RNN model trained only on linear and step strain inputs can perform very well on prediction of completely different quadratic strain inputs, demonstrating certain level of generalization ability in extrapolation. Moreover, it is observed that the extrapolation ability depends on the types of strain inputs. The performance is better for continuous strain inputs than that for jump strain inputs. The differences in the generalization ability of RNN models in viscoelasticity and other history-dependent materials are discussed. It suggests that RNN data-driven modeling can be an alternative to the conventional viscoelasticity models. [ABSTRACT FROM AUTHOR]