Taylor bubble velocity in slug flow is a closure relation which significantly affects the prediction of liquid holdup (or void fraction) and pressure gradient in mechanistic models of slug flow for ...oil and gas pipe applications. In this work, we use a validated Computational Fluid Dynamics (CFD) approach to simulate the motion of Taylor bubbles in pipes; the interface is tracked with a Level-Set method implemented in a commercial code. A large numerical database is generated covering the most ample range of fluid properties and pipe inclination angles explored to date (Eo ∊ 10, 700,Mo ∊ 1 x 10-6, 5 x 103, and θ ∊ 0°, 90°). A unified Taylor bubble rise velocity correlation is extracted from that database. The new correlation predicts the numerical database with 8.6% absolute average relative error and a coefficient of determination R2 = 0.97 and other available experimental data with 13.0% absolute average relative error and R2 = 0.84 outperforming existing correlations and models.
Meeting the Contact-Mechanics Challenge Müser, Martin H.; Dapp, Wolf B.; Bugnicourt, Romain ...
Tribology letters,
12/2017, Letnik:
65, Številka:
4
Journal Article
Recenzirano
Odprti dostop
This paper summarizes the submissions to a recently announced contact-mechanics modeling challenge. The task was to solve a typical, albeit mathematically fully defined problem on the adhesion ...between nominally flat surfaces. The surface topography of the rough, rigid substrate, the elastic properties of the indenter, as well as the short-range adhesion between indenter and substrate, were specified so that diverse quantities of interest, e.g., the distribution of interfacial stresses at a given load or the mean gap as a function of load, could be computed and compared to a reference solution. Many different solution strategies were pursued, ranging from traditional asperity-based models via Persson theory and brute-force computational approaches, to real-laboratory experiments and all-atom molecular dynamics simulations of a model, in which the original assignment was scaled down to the atomistic scale. While each submission contained satisfying answers for at least a subset of the posed questions, efficiency, versatility, and accuracy differed between methods, the more precise methods being, in general, computationally more complex. The aim of this paper is to provide both theorists and experimentalists with benchmarks to decide which method is the most appropriate for a particular application and to gauge the errors associated with each one.
Machine Learning methods and, in particular, Artificial Neural Networks (ANNs) have demonstrated promising capabilities in material constitutive modeling. One of the main drawbacks of such approaches ...is the lack of a rigorous frame based on the laws of physics. This may render physically inconsistent the predictions of a trained network, which can be even dangerous for real applications.
Here we propose a new class of data-driven, physics-based, neural networks for constitutive modeling of strain rate independent processes at the material point level, which we define as Thermodynamics-based Artificial Neural Networks (TANNs). The two basic principles of thermodynamics are encoded in the network’s architecture by taking advantage of automatic differentiation to compute the numerical derivatives of a network with respect to its inputs. In this way, derivatives of the free-energy, the dissipation rate and their relation with the stress and internal state variables are hardwired in the architecture of TANNs. Consequently, our approach does not have to identify the underlying pattern of thermodynamic laws during training, reducing the need of large data-sets. Moreover the training is more efficient and robust, and the predictions more accurate. Finally and more important, the predictions remain thermodynamically consistent, even for unseen data. Based on these features, TANNs are a starting point for data-driven, physics-based constitutive modeling with neural networks.
We demonstrate the wide applicability of TANNs for modeling elasto-plastic materials, using both hyper- and hypo-plasticity models. Strain hardening and softening are also considered for the hyper-plastic scenario. Detailed comparisons show that the predictions of TANNs outperform those of standard ANNs. Finally, we demonstrate that the implementation of the laws of thermodynamics confers to TANNs high robustness in the presence of noise in the training data, compared to standard approaches.
TANNs’ architecture is general, enabling applications to materials with different or more complex behavior, without any modification.
Display omitted
•A new class of data-driven, physics-based, neural networks is proposed for constitutive modeling of strain rate independent processes at the material point level.•The two basic principles of thermodynamics are hardwired in the architecture of Thermodynamics-based Artificial Neural Networks.•The proposed approach is applied for modeling elasto-plastic materials, using both hyper- and hypo-plasticity models, with strain softening and hardening.•Detailed comparisons show that the predictions of the proposed class outperform those of standard neural network approaches.•The implementation of the laws of thermodynamics is found to confer to TANNs high robustness to noise and improved accuracy.