Abstract: Multiphase fluid flow with complex compositions is an increasingly attractive research topic with more and more attentions paid on related engineering problems, including global warming and green house effect, oil recovery enhancement and subsurface water pollution treatment. Prior to study the flow behaviors and phase transitions in multi-component multiphase flow, the first effort should be focused on the accurate prediction of the total phase numbers existing in the fluid mixture, and then the phase equilibrium status can be determined. In this paper, a novel and fast prediction technique is proposed based on deep learning method. The training data is generated using a selected VT dynamic flash calculation scheme and the network constructions are deeply optimized on the activation functions. Compared to previous machine learning techniques proposed in literatures to accelerate vapor liquid phase equilibrium calculation, the total number of phases existing in the mixture is determined first and other phase equilibrium properties will be estimated then, so that we do not need to ensure that the mixture is in two phase conditions any more. Our method could handle fluid mixtures with complex compositions, with 8 different components in our example and the original data is in a large amount. The analysis on prediction performance of different deep learning models with various neural networks using different activation functions can help future researches selecting the features to construct the neural network for similar engineering problems. Some conclusions and remarks are presented at the end to help readers catch our main contribution and insight the future related research.

Abstract: Modeling and simulation of fluid flow and heat transfer processes occurring in heterogeneous irregular regions have received extensive attention in recent years. The presence of heterogeneous properties would exert crucial impacts on the overall performance of fluid flow and heat transfer simulations. For example, the high heterogeneous properties always worsen the model coefficient matrix and complicate the simulation difficulty. Therefore, the need to develop high-efficient and accurate numerical methods for general fluid flow and heat transfer occurring in heterogeneous irregular regions which could significantly reduce the computational efforts at the same time conserve the main physical properties, is highly addressed among engineering and academic communities. In this study, we present a highly efficient solver, geometrical multi-grid (GMG), for the fast simulation of fluid flow and heat transfer problems occurring in heterogeneous irregular regions in the framework of body-fitted coordinate (BFC) system. The key point of the proposed multigrid solver lies in the calculation of heterogeneous properties on coarse grid levels within the original physical domain, in which the up-scaling method is widely used. However, different upscaling methods would yield the effective properties with different numerical accuracy and computational efficiency. To explore the influence of the upscaling approaches on overall performances of the proposed multigrid solver, in this study we adopt the general statistical averages (e.g. harmonic average, arithmetic average, geometric average, harmonic-arithmetic average, etc.) and flow-based methods (e.g. sealed-side boundary condition, open-side boundary conditions, etc. for fluid flows) to compute the unscaled effective properties on corresponding coarse grid levels. The numerical accuracy of the quantity of interest on different grid levels and the computational speed-up of the proposed multigrid solver for flow and heat transfer problems occurring in heterogeneous irregular regions are validated by several examples to assess the influence of different upscaling approaches on computations. The proposed multigrid solver for fluid flow and heat transfer problems in heterogeneous irregular regions can not only markedly improve the computational efficiency of the fine grid solution, but also can provide the computation byproduct - solution on coarse grid levels for specific applications, for example in which the coarse grid solution can be used for sample recycling in multigrid multilevel Monte Carlo method to avoid the repeated realization of sampling on coarse grid levels.

Abstract: Enhanced geothermal systems (EGS) are the major way of the hot dry rock (HDR) exploitation. At present, the finite element method (FEM) is often used to simulate the thermal energy extraction process of the EGS. Satisfactory results can be obtained by this method to a certain extent. However, when many discrete fractures exist in the computational domain, a large number of unstructured grids must be used, which seriously affects the computational efficiency. To solve this challenge, based on the embedded discrete fracture model (EDFM), two sets of seepage and energy conservation equations are respectively used to describe the flow and heat transfer processes of the matrix media and the fracture media. The main advantage of the proposed model is that the structured grid can be used to mesh the matrix, and there is no need to refine the mesh near the fracture. Comparing with commercial software, COMSOL Multiphysics, the accuracy of the proposed model is verified. Subsequently, a specific example of geothermal exploitation is designed, and the spatial-temporal evolutions of pressure and temperature fields are analyzed.

Abstract: Modelling the fusion and heat affected microstructure of an Additive Manufacturing (AM) process bridges many length and time scales and requires more than intelligent meshing schemes to make simulations feasible. The aim of this research was to develop an efficient and simple, yet significantly accurate high quality and high precision thermal model in wire + arc additive manufacturing process. To describe the influence of the process parameters and materials on the entire welding process, a 3D transient non-linear finite element model to simulate multi-layer deposition of cast IN-738LC alloy onto SAE-AISI 1524 Carbon Steel Substrates was developed. Temperature-dependent material properties and the effect of forced convection were included in the model. The heat source represented by a moving Gaussian power density distribution was applied over the top surface of the specimen during a period of time that depends on the welding speed. The effect of multi-layer deposition on the prediction and validation of melting pool shape and thermal cycles was also investigated. The effect of convection and radiation heat loss from the weldment (layers) surfaces were included into the finite element analysis. As the AM layers itself act as extended surfaces (fins), it was found that the heat extraction is quite significant. It is encouraging to note that the thermal model is sufficiently accurate to predict thermal cycles, FZ and HAZ weld profiles. A firm foundation for modelling thermal transport in wire + arc additive manufacturing process it was established.