In the fields of physics and engineering, it is crucial to understand phase transition dynamics. This field involves fundamental partial differential equations (PDEs) such as the Allen–Cahn, Burgers, and two-dimensional (2D) wave equations. In alloys, the evolution of the phase transition interface is described by the Allen–Cahn equation. Vibrational and wave phenomena during phase transitions are modeled using the Burgers and 2D wave equations. The combination of these equations gives comprehensive information about the dynamic behavior during a phase transition. Numerical modeling methods such as finite difference method (FDM), finite volume method (FVM) and finite element method (FEM) are often applied to solve phase transition problems that involve many partial differential equations (PDEs). However, physical problems can lead to computational complexity, increasing computational costs dramatically. Physics-informed neural networks (PINNs), as new neural network algorithms, can integrate physical law constraints with neural network algorithms to solve partial differential equations (PDEs), providing a new way to solve PDEs in addition to the traditional numerical modeling methods. In this paper, a hard-constraint wide-body PINN (HWPINN) model based on PINN is proposed. This model improves the effectiveness of the approximation by adding a wide-body structure to the approximation neural network part of the PINN architecture. A hard constraint is used in the physically driven part instead of the traditional practice of PINN constituting a residual network with boundary or initial conditions. The high accuracy of HWPINN for solving PDEs is verified through numerical experiments. One-dimensional (1D) Allen–Cahn, one-dimensional Burgers, and two-dimensional wave equation cases are set up for numerical experiments. The properties of the HWPINN model are inferred from the experimental data. The solution predicted by the model is compared with the FDM solution for evaluating the experimental error in the numerical experiments. HWPINN shows great potential for solving the PDE forward problem and provides a new approach for solving PDEs.

The traditional multi-level grid multiple-relaxation-time lattice Boltzmann method (MRT-LBM) requires interpolation calculations in time and space. It is a complex and computationally intensive process. By using the buffer technique, this paper proposes a new multi-level grid MRT-LBM which requires only spatial interpolation calculations. The proposed method uses a center point format to store multi-level grid information. The grid type determination in the flow field calculation domain is done using the axis aligned bounding box (AABB) triangle overlap test. According to the calculation characteristics of MRT-LBM, the buffer grid is proposed for the first time at the interface of different levels of grids, which is used to remove the temporal interpolation calculation and simplify the spatial interpolation calculation. The corresponding multi-level grid MRT-LBM algorithm is also presented for two-dimensional and three-dimensional flow field calculation problems. For the two-dimensional problem of flow around a circular cylinder, the simulation results show that a four-level grid MRT-LBM proposed in this paper can accurately obtain the aerodynamic coefficients and Strouhal number at different Reynolds numbers, and it has about 1/9 of the total number of grids as a single-level grid MRT-LBM and is 6.76 times faster. For the three-dimensional flow calculation problem, the numerical experiments of flow past a sphere are simulated to verify the numerical precision of the presented method at Reynolds numbers = 100, 200, 250, 300, and 1000. With the streamlines and velocity contours, it is demonstrated that the multi-level grid MRT-LBM can be calculated accurately even at the interface of different size grids.

Physics-Informed Neural Networks (PINNs) improve the efficiency of data utilization by combining physical principles with neural network algorithms and thus ensure that their predictions are consistent and stable with the physical laws. PINNs open up a new approach to address inverse problems in fluid mechanics. Based on the single-relaxation-time lattice Boltzmann method (SRT-LBM) with the Bhatnagar–Gross–Krook (BGK) collision operator, the PINN-SRT-LBM model is proposed in this paper for solving the inverse problem in fluid mechanics. The PINN-SRT-LBM model consists of three components. The first component involves a deep neural network that predicts equilibrium control equations in different discrete velocity directions within the SRT-LBM. The second component employs another deep neural network to predict non-equilibrium control equations, enabling the inference of the fluid’s non-equilibrium characteristics. The third component, a physics-informed function, translates the outputs of the first two networks into physical information. By minimizing the residuals of the physical partial differential equations (PDEs), the physics-informed function infers relevant macroscopic quantities of the flow. The model evolves two sub-models that are applicable to different dimensions, named the PINN-SRT-LBM-I and PINN-SRT-LBM-II models according to the construction of the physics-informed function. The innovation of this work is the introduction of SRT-LBM and discrete velocity models as physical drivers into a neural network through the interpretation function. Therefore, the PINN-SRT-LBM allows a given neural network to handle inverse problems of various dimensions and focus on problem-specific solving. Our experimental results confirm the accurate prediction by this model of flow information at different Reynolds numbers within the computational domain. Relying on the PINN-SRT-LBM models, inverse problems in fluid mechanics can be solved efficiently.

Mesh quality is a major factor affecting the structure of computational fluid dynamics (CFD) calculations. Traditional mesh quality evaluation is based on the geometric factors of the mesh cells and does not effectively take into account the defects caused by the integrity of the mesh. Ensuring the generated meshes are of sufficient quality for numerical simulation requires considerable intervention by CFD professionals. In this paper, a Transformer-based network for automatic mesh quality evaluation (Gridformer), which translates the mesh quality evaluation into an image classification problem, is proposed. By comparing different mesh features, we selected the three features that highly influence mesh quality, providing reliability and interpretability for feature extraction work. To validate the effectiveness of Gridformer, we conduct experiments on the NACA-Market dataset. The experimental results demonstrate that Gridformer can automatically identify mesh integrity quality defects and has advantages in computational efficiency and prediction accuracy compared to widely used neural networks. Furthermore, a complete workflow for automatic generation of high-quality meshes based on Gridformer was established to facilitate automated mesh generation. This workflow can produce a high-quality mesh with a low-quality mesh input through automatic evaluation and optimization cycles. The preliminary implementation of automated mesh generation proves the versatility of Gridformer.

As an important economic fish resource, germplasm resources and phenotypic measurements of tilapia are of great importance in the direction of culture and genetic improvement. Furthermore, accurate identification and precise localization of tilapia body parts are crucial for enabling key technologies such as automated capture and precise cutting. However, there are some problems in the semantic segmentation of tilapia fish, including the accuracy of target edge segmentation and the ambiguity in segmenting small targets. To improve the accuracy of semantic segmentation of tilapia parts in real farming environments, an improved Deeplabv3+ network model method is proposed for implementing tilapia part segmentation to facilitate phenotypic measurements on tilapia in this paper. The CBAM module is embedded in the encoder, which can improve the accurate identification and localization of tilapia parts by adaptively adjusting the channel weights and spatial weights and better focus on the key features and spatial connections of tilapia body parts. Furthermore, the decoding part of the Deeplabv3+ model is optimized by using SENet, which greatly increases the segmentation accuracy of the network by establishing the interdependence between channels while suppressing useless features. Finally, model performance is tested and compared with the original network and other methods on the tilapia part segmentation dataset. The experimental results show that the segmentation performance of the improved network is better than other networks, such as PSPNet and U-Net, and the IoU values in the head, fins, trunk, and tail of the fish body are 9.78, 2.27, 6.27, and 6.58 percentage points higher than those of the Deeplabv3+ network, respectively. The results validate the effectiveness of our approach in solving the above problems encountered in the semantic segmentation of tilapia parts.

As an improved method of the lattice Boltzmann method (LBM), the regularized lattice Boltzmann method (RLBM) has been widely used to simulate fluid flow. For solving high Reynolds number problems, large eddy simulation (LES) and RLBM can be combined. The computation of fluid flow problems often requires a large number of computational grids and large-scale parallel clusters. Therefore, the high scalability parallel algorithm of RLBM with LES on a large-scale cluster has been proposed in this paper. The proposed parallel algorithm can solve complex flow problems with large-scale Cartesian grids and high Reynolds numbers. In order to achieve computational load balancing, the domain decomposition method (DDM) has been used in large-scale mesh generation. Three mesh generation strategies are adopted, namely 1D, 2D and 3D. Then, the buffer on the grid interface is introduced and the corresponding 1D, 2D and 3D parallel data exchange strategies are proposed. For the 3D lid-driven cavity flow and incompressible flow around a sphere under a high Reynolds number, the given parallel algorithm is analyzed in detail. Experimental results show that the proposed parallel algorithm has a high scalability and accuracy on hundreds of thousands of cores.