Traditional Numerical Simulation Research Articles

In Silico Design of Freeform Solar Cell Structures from High‐Throughput Artificial Intelligence‐Generated Configurations

Recent advances in artificial intelligence‐generated configurations (AIGC) have transformed various fields in science and technology domains and enabled the inverse design of materials and structures for enhanced performance with minimal human input. The patterned micro‐/nanostructures are widely adopted to increase power conversion efficiencies in thin‐film organic solar cells (OSCs), making these structures highly suitable for AIGC applications. Although the computational cost of traditional numerical simulations is a barrier to making AIGC high throughput, this issue is addressed by integrating the high‐speed automated machine learning (AutoML) Analyzer with genetic algorithm‐based topology optimization algorithms. Compared to standard numerical solutions, the AutoML Analyzer in the proposed system predicts outcomes ≈22 700 times faster, resulting in a 98.47% reduction in computation costs, while maintaining high average accuracies of ≈99%. By evaluating an extensive dataset of 600 000 AIGCs, an optimized PM6:Y6 OSC device design with a power conversion efficiency of 17.58% is identified, significantly outperforming the 14.70% baseline efficiency of the OSC device without the poly methyl methacrylate scattering and substrate layers. Results underscore the potential of AIGC techniques for efficiently enhancing the performance of photovoltaic devices in renewable energy applications.

Numerical simulation on diffusion–reaction behavior of concrete under sulfate–chloride coupled attack

The transport of sulfate–chloride salts in concrete is the main cause of its structural deterioration. Traditional numerical simulation assumes that the diffusion coefficients of sulfate ions and chloride ions are both constant leading to large deviations in the predicted results. Based on the mass conservation law, Fick’s second law, and the theory of porous media, a variable coefficient transport model of sulfate–chloride salts in concrete is established in this paper. The model fully reflects the adsorption and desorption of chloride ions in the transport process, the reaction-filling process of sulfate ions and hydration products, as well as the most important porosity and tortuosity affecting ion transport. Results show that the prediction results of the two-dimensional variable coefficient model are consistent with experimental and literature results, with maximum error of 23.49%. The analysis of influencing factors shows that the interaction between sulfate ions and chloride ions does not change the transport law of ions in concrete, but delays the rate of ion corrosion within about 700 days. The ability of chloride ions to be adsorbed decreases by about 20% when sulfate ions are present, and the desorption rate is about 40% to 60%. The tortuosity decreases with the increase of water-binder ratio, and increasing the initial porosity from 5% to 12% leads to an early appearance of cracks by about 1400 days.

3D reconstruction of digital rock guided by petrophysical parameters with deep learning

Digital rock reconstruction is the basis of digital petrophysics, aiming to provide diverse samples for heterogeneity analysis on microscopic pore structures. Traditional physical experiments and numerical simulation methods tend to lead to a high experimental cost or a long reconstruction time, which is improved by generative adversarial networks (GAN) in recent studies. However, GAN's reconstructions are still too stochastic to satisfy the need for joint control over user-defined feature combinations. Therefore, we propose a multi-conditional GAN (MCGAN) model, using petrophysical parameters to, in turn, guide and drive the training process. Prior information, such as porosity, tortuosity, specific surface area, and fractal dimension, is concatenated with random noise to precisely control the pore distribution of dummy samples. In the loss function, we design a feature loss term to accelerate convergence and allow users the freedom to choose parameter combinations requiring tighter constraints. We compare MCGAN with three other GAN-based models on petrophysical properties, statistical features, and visual effects for performance evaluation. The results show that for applicable parameters, our method supports better mapping performance and higher degrees of freedom in multi-conditional reconstruction of digital rock. Therefore, MCGAN can quickly reconstruct digital cores with specific feature combinations according to users' needs. This work proposes a generative model for binary digital rock based on deep learning, promising to help more precisely analyze the microscopic heterogeneity of pore structures.

An adaptive hybrid grids finite-element approach for plane wave three-dimensional electromagnetic modeling

In unstructured finite element (FE) plane wave electromagnetic modeling, the goal-oriented mesh adaptive method based on posterior errors is often used to optimize the mesh to a state where a high-precision numerical solution can be obtained with fewer elements. However, for high-frequency data, this method tends to generate a large number of small elements in the shallow near-surface area, which increases the computational burden of FE solution. To solve this problem, we developed an improved goal-oriented hybrid mesh adaptive method. This method realizes the coupling of prismatic elements and tetrahedral elements in FE analysis and can establish FE system equations without suspension nodes. Numerical example studies illustrate how the proposed method uses triangular prism elements to divide the near-surface area and greatly improves the computational efficiency of plane wave electromagnetic modeling. For high-frequency modeling, the proposed hybrid grid reduces the number of required elements, improves the numerical accuracy, reduces the iterations of adaptive grids refinement, and improves the efficiency of the adaptive numerical solution. For low-frequency forward calculations, the hybrid grid maintains the same solution efficiency as traditional adaptive numerical simulation based on pure tetrahedrons.

Application of Machine Learning in Material Synthesis and Property Prediction.

Material innovation plays a very important role in technological progress and industrial development. Traditional experimental exploration and numerical simulation often require considerable time and resources. A new approach is urgently needed to accelerate the discovery and exploration of new materials. Machine learning can greatly reduce computational costs, shorten the development cycle, and improve computational accuracy. It has become one of the most promising research approaches in the process of novel material screening and material property prediction. In recent years, machine learning has been widely used in many fields of research, such as superconductivity, thermoelectrics, photovoltaics, catalysis, and high-entropy alloys. In this review, the basic principles of machine learning are briefly outlined. Several commonly used algorithms in machine learning models and their primary applications are then introduced. The research progress of machine learning in predicting material properties and guiding material synthesis is discussed. Finally, a future outlook on machine learning in the materials science field is presented.

An enhanced prediction framework for coalbed methane production incorporating deep learning and transfer learning

Coalbed methane (CBM) is one of the most promising clean energy sources, and its production forecast is of vital strategic significance for the sustainable development. Traditional numerical simulation encounters the curse of dimensionality when applied to multi-scale reservoir modeling. Conventional machine learning faces challenges in modeling under small-sample condition. This study leverages a bidirectional long short-term memory (Bi-LSTM) to better capture the complex dynamic data characteristics of CBM. Moreover, we incorporate transfer learning to generalize the implicit laws learned from 1205 vertical wells to the modeling of horizontal wells, establishing an enhanced prediction framework that can be adapted to different well types. The findings confirm that Bi-LSTM has more powerful memory capability for CBM dynamic data with prediction errors below 200 m³, which can improve the prediction accuracy by about 15% compared to the traditional methods. The fusion of transfer learning can significantly enhance the prediction performance by approximately 10% for both low-producing, medium-producing, and high-producing horizontal wells, illustrating that the dynamic data of vertical wells located in the same block contain implicit features that are crucial guides to the modeling of horizontal wells. Notably, the enhanced architecture exhibits greater flexibility and adaptability under small-sample condition for complex production patterns.

Multilevel method for predicting flow fields in radial turbines based on sparsity-promoting dynamic mode decomposition

PurposeThe purpose of this study is to propose a fast method for predicting flow fields with periodic behavior with verification in the context of a radial turbine to meet the urgent requirement to effectively capture the unsteady flow characteristics in turbomachinery. Aiming at meeting the urgent requirement to effectively capture the unsteady flow characteristics in turbomachinery, a fast method for predicting flow fields with periodic behavior is proposed here, with verification in the context of a radial turbine (RT).Design/methodology/approachSparsity-promoting dynamic mode decomposition is used to determine the dominant coherent structures of the unsteady flow for mode selection, and for flow-field prediction, the characteristic parameters including amplitude and frequency are predicted using one-dimensional Gaussian fitting with flow rate and two-dimensional triangulation-based cubic interpolation with both flow rate and rotation speed. The flow field can be rebuilt using the predicted characteristic parameters and the chosen model.FindingsUnder single flow-rate variation conditions, the turbine flow field can be recovered using the first seven modes and fitted amplitude modulus and frequency with less than 5% error in the pressure field and less than 9.7% error in the velocity field. For the operating conditions with concurrent flow-rate and rotation-speed fluctuations, the relative error in the anticipated pressure field is likewise within an acceptable range. Compared to traditional numerical simulations, the method requires a lot less time while maintaining the accuracy of the prediction.Research limitations/implicationsIt would be challenging and interesting work to extend the current method to nonlinear problems.Practical implicationsThe method presented herein provides an effective solution for the fast prediction of unsteady flow fields in the design of turbomachinery.Originality/valueA flow prediction method based on sparsity-promoting dynamic mode decomposition was proposed and applied into a RT to predict the flow field under various operating conditions (both rotation speed and flow rate change) with reasonable prediction accuracy. Compared with numerical calculations or experiments, the proposed method can greatly reduce time and resource consumption for flow field visualization at design stage. Most of the physics information of the unsteady flow was maintained by reconstructing the flow modes in the prediction method, which may contribute to a deeper understanding of physical mechanisms.

A computational method for complex-shaped hydraulic turbomachinery flow based on the immersed boundary method

Traditional numerical simulation techniques, such as sliding mesh, dynamic mesh, and others, have many limitations in dealing with flow simulation with the large-scale movement of solid boundaries, which is the case for simulating the flow of complex-shaped hydraulic turbomachinery such as propellers, pumps, and turbines. The immersed boundary (IB) method provides a new approach to solve the above-mentioned limitations. Therefore, this study proposes a sharp-interface IB method based on the level-set function that is suitable for simulating the flow through turbomachinery with complex geometries. This method is applied to actual three-dimensional numerical simulations of high-Reynolds number propellers using an in-house computational fluid dynamics solver. The results show that the proposed method can provide comparatively accurate predictions of unsteady load coefficients within the propeller flow passage and capture the correct propeller wake characteristics as well as the interaction between the propeller wake and free surface. This study is aimed at providing a theoretical basis and engineering reference for the application of the IB method in engineering numerical simulations.

Explosion of high pressure hydrogen tank in fire: Mechanism, criterion, and consequence assessment

This study published experimental data on the catastrophic rupture consequences of high-pressure hydrogen storage tanks in fire environments. It made up for the lack of actual explosion data for verification of the traditional theoretical prediction models and numerical simulations. To reveal the mechanism of high-pressure tank explosion and corresponding characteristics of hydrogen explosion, the fireball parameters, fragment characteristics, and blast wave attenuation patterns resulting from tank rupture were analyzed. The results indicated that the hazard of hydrogen storage tank explosion was coupled with the combined contribution of physical and chemical explosion energies. The failure pressure of a 6.8 L – 30 MPa tank under fire conditions decreased by 60.3 % compared to that at room temperature. Following tank explosion, the duration of hydrogen-air deflagration was only about 2 s, and the maximum diameter of the fireball was nearly 4.48 m. A portion of the mechanical energy generated by tank explosion was converted into the kinetic energy of projectile fragments, with the farthest discovered fragment distance reaching 46.0 m. Additionally, the measured peak overpressure decreased from 875.33 kPa to 7.52 kPa at distances ranging from 2 m to 15 m from the explosion source. To evaluate blast wave overpressure in high-pressure hydrogen tanks for engineering applications, a standardized and generalized procedure was innovatively proposed, and its effectiveness was validated by experimental data. Furthermore, the hazardous distances defined by the peak overpressure were utilized to delineate risk zones. The distances of 61.6 m and 17.3 m from the explosion source were identified as zones without injury to human and without damage to structures, respectively. The proposed methodology and the results of the experiments were of practical importance for ensuring the safety of high-pressure hydrogen storage systems in various applications.

Intelligent Optimization of Gas Flooding Based on Multi-Objective Approach for Efficient Reservoir Management

The efficient development of oil reservoirs mainly depends on the comprehensive optimization of the subsurface fluid flow process. As an intelligent analysis technique, artificial intelligence provides a novel solution to multi-objective optimization (MOO) problems. In this study, an intelligent agent model based on the Transformer framework with the assistance of the multi-objective particle swarm optimization (MOPSO) algorithm has been utilized to optimize the gas flooding injection–production parameters in a well pattern in the Middle East. Firstly, 10 types of surveillance data covering 12 years from the target reservoir were gathered to provide a data foundation for model training and analysis. The prediction performance of the Transformer model reflected its higher accuracy compared to traditional reservoir numerical simulation (RNS) and other intelligent methods. The production prediction results based on the Transformer model were 21, 12, and 4 percentage points higher than those of RNS, bagging, and the bi-directional gated recurrent unit (Bi-GRU) in terms of accuracy, and it showed similar trends in the gas–oil ratio (GOR) prediction results. Secondly, the Pareto-based MOPSO algorithm was utilized to fulfil the two contradictory objectives of maximizing oil production and minimizing GOR simultaneously. After 10,000 iterations, the optimal injection–production parameters were proposed based on the generated Pareto frontier. To validate the feasibility and superiority of the developed approach, the development effects of three injection–production schemes were predicted in the intelligent agent model. In the next 400 days of production, the cumulative oil production increased by 25.3% compared to the average distribution method and 12.7% compared to the reservoir engineering method, while GOR was reduced by 27.1% and 15.3%, respectively. The results show that MOPSO results in a strategy that more appropriately optimizes oil production and GOR compared to some previous efforts published in the literature. The injection–production parameter optimization method based on the intelligent agent model and MOPSO algorithm can help decision makers to update the conservative development strategy and improve the development effect.

Magnetotelluric Deep Learning Forward Modeling and Its Application in Inversion

Magnetotelluric (MT) inversion and forward modeling are closely linked. The optimization and iteration processes of the inverse algorithm require frequent calls to forward modeling. However, traditional numerical simulations for forward modeling are computationally expensive; here, deep learning (DL) networks can simulate forward modeling and significantly improve forward speed. Applying DL for forward modeling in inversion problems requires a high-precision network capable of responding to fine changes in the model to achieve high accuracy in inversion optimization. Most existing MT studies have used a convolutional neural network, but this method is limited by the receptive field and cannot extract global feature information. In contrast, the Mix Transformer has the ability to globally model and extract features. In this study, we used a Mix Transformer to hierarchically extract feature information, adopted a multiscale approach to restore feature information to the decoder, and eliminated the skip connection between the encoder and decoder. We designed a forward modeling network model (MT-MitNet) oriented toward inversion. A sample dataset required for DL forward was established using the forward data generated from the traditional inverse calculation iteration process. The trained network quickly and accurately calculates the forward response. The experimental results indicate a high agreement between the forward results of MT-MitNet and those obtained with traditional methods. When MT-MitNet replaces the forward computation in traditional inversion, the inversion results obtained with it are also highly in agreement with the traditional inversion results. Importantly, under the premise of ensuring high accuracy, the forward speed of MT-MitNet is hundreds of times faster than that of traditional inversion methods in the same process.

Performance evaluation of a single footing over rigid inclusions-reinforced soft soil - use of a novel LWSO-RF model to predict load transfer efficiency

The rigid inclusion (RI) reinforcement technique has been increasingly used to reduce settlements for engineering constructions over soft soils. However, the load transfer mechanisms within the granular mattress are not well understood, especially for the RI-reinforced systems with the presence of structural elements such as footings. To evaluate the performance of a single footing over RI-reinforced soft soils, a novel intelligence approach combining the white shark optimizer (WSO) and the Logistic chaotic mapping (LCM) was proposed to optimize the random forest (RF) model for predicting the load transfer efficiency (LTE). Seven variables were considered to systematically investigate the LTE, including the loading magnitude (L), the load eccentricity (LE), the geometrical and strength parameters of the load transfer platform thickness (LT), Young’s modulus (E), cohesion (C) and friction angle (F)) and the compression ratio (CR). Based on the feature selection method, nine different combination models were developed to estimate the LTE. Their performances were compared using the determination coefficient (R2), the root mean squared error (RMSE), the variance accounted for (VAF) and the Willmott’s index (WI). The evaluation results indicated that the LWSO-RF model containing seven variables (L, LT, LE, E, CF, FA, CR) is the best one as it gives the highest prediction accuracy in both the training and testing phases. The sensitivity results demonstrated that LT is the most important variable for the LTE prediction. This work utilizes an intelligence approach to extend the traditional numerical simulation analysis for predicting the LTE, which provides additional guidance to further assess the performance of foundations over RI-reinforced soft soil.

Research on status‐driven hybrid evaluation method for mission success probability

AbstractMission Success Probability is a top‐level indicator that integrates reliability, maintainability, and spare parts security during the system's mission. Numerical simulation is used for mission success evaluation of complex system because it can consider any form of system structure and component failure distribution model. In actual engineering, system components usually deteriorate or fail with the extension of service time and need to be repaired or replaced, so there are dynamic changes in component status during the life cycle. To evaluate the mission success in such dynamic scenarios, it is necessary to repeat the simulation based on the whole system once the local parameters are adjusted. Despite the continuous improvement in computational performance, traditional numerical simulations are not well matched for rapid evaluation and timely decision making in dynamic scenarios. Therefore, an evaluation method that mixes proxy model and simulation sampling is proposed in this paper. The method uses the simulation to complete the data acquisition, and then analyzes the data by inverse function sampling and Copula function to obtain the model training samples, and finally realizes the approximate description of the local unit simulation process by proxy model. The experimental results show that the method can reduce the repetitive simulation scale from the whole system to local unit while ensuring high accuracy, avoiding many repeated calculations in variable scenarios.

High-accuracy approximation of evolutionary pairwise games on complex networks

Previous studies have shown that the topological properties of a complex network, such as heterogeneity and average degree, affect the evolutionary game dynamics on it. However, traditional numerical simulations are usually time-consuming and demand a lot of computational resources. In this paper, we propose the method of dynamical approximate master equations (DAMEs) to accurately predict the evolutionary outcomes on complex networks. We demonstrate that the accuracy of DAMEs supersedes previous standard pairwise approximation methods, and DAMEs require far fewer computational resources than traditional numerical simulations. We apply the DAMES to investigate prisoner’s dilemma and snowdrift game on regular and scale-free networks, demonstrating its effectiveness. Overall, our method facilitates the investigation of evolutionary dynamics on a broad range of complex networks, and provides new insights into the puzzle of cooperation.

Deep learning method for super-resolution reconstruction of the spatio-temporal flow field

The high-resolution (HR) spatio-temporal flow field plays a decisive role in describing the details of the flow field. In the acquisition of the HR flow field, traditional direct numerical simulation (DNS) and other methods face a seriously high computational burden. To address this deficiency, we propose a novel multi-scale temporal path UNet (MST-UNet) model to reconstruct temporal and spatial HR flow fields from low-resolution (LR) flow field data. Different from the previous super-resolution (SR) model, which only takes advantage of LR flow field data at instantaneous (SLR) or in a time-series (MTLR), MST-UNet introduces multi-scale information in both time and space. MST-UNet takes the LR data at the current frame and the predicted HR result at the previous moment as the model input to complete the spatial SR reconstruction. On this basis, a temporal model is introduced as the inbetweening model to obtain HR flow field data in space and time to complete spatio-temporal SR reconstruction. Finally, the proposed model is validated by the spatio-temporal SR task of the flow field around two-dimensional cylinders. Experimental results show that the outcome of the MST-UNet model in spatial SR tasks is much better than those of SLR and MTLR, which can greatly improve prediction accuracy. In addition, for the spatio-temporal SR task, the spatio-temporal HR flow field predicted by the MST-UNet model has higher accuracy either.

Numerical Simulation of the Lubricant-Solid Interface Using the Multigrid Method

Solid asperity interactions are common and inevitable under severe loading conditions for any lubricated contact. Heavy-duty machine components (gears, bearings, etc.) generally operate under Mixed Lubrication (ML), where uneven surface features contact each other when the generated fluid pressure is not enough to support the external load. The Reynolds equation is commonly used to simulate smooth lubricated contacts numerically. In rough lubricated interfaces where opposite surface asperities make contact, the Reynolds equation alone cannot accurately predict pressure using the traditional numerical simulation method. In this paper, lubrication–contact interface conditions (LCICs) have been implemented and extended to solve the multiple asperity contact problem using the full-multigrid approach. The developed novel algorithm has significantly accelerated the solution process and improved the accuracy and efficiency of pressure calculation for fluid–solid sub-interactions that can occur in ML regions. The results of the finite difference method (FDM) results have been compared with those of computational fluid dynamics (CFD) simulation to validate the newly developed model. Hence, the proposed optimized solution method will provide valuable insight to researchers and industry engineers interested in simulating the ML problem where the effect of the fluid–solid interface can be captured effectively to improve reliability in the calculation of the life expectancy of the lubricated parts.

Study on boundary reflection mechanism of continuous wave mud pulse in the flow loop

Continuous wave (CW) mud pulse is one of the essential ways of high speed data transmission while drilling. The boundary reflection process and mechanism need to be clarified, which seriously affects the analog interpretation of pressure waves in flow loops and drilling conditions. In this paper, the boundary reflection of mass inlet and pressure outlet is studied by using traditional acoustic theory and numerical simulation methods, and mutual verification is carried out with flow loop. The results show that the pressure wave generated by the CW generator can be divided into compression wave and expansion wave and transmitted to the upstream and downstream pipelines. The inlet boundary reflection is the hard boundary reflection in acoustics. The phase of reflected wave is the same as that of the incident wave. The outlet boundary is the soft boundary, the reflected wave phase is 180° opposite to the incident wave. In particular, the process and mechanism of secondary reflection of stator and rotor structure on pressure wave are analyzed for the first time, which provides a theoretical basis for pressure wave amplitude. Finally, due to the possibility of cavitation phenomenon in the flow loop, it is different from the pressure environment of the bottom under drilling conditions and should to study further.

Novel method for obtaining transfer characteristics of subway-induced ground vibrations

Aiming at the multiple-input multiple-output vibration problem in rail transit, this paper presents a novel method for obtaining transfer characteristics of vehicle-induced vibrations. To verify its accuracy and applicability, the obtained transfer coefficients are further combined with vehicle-track coupling simulations to predict environmental vibrations under different conditions. The results prove that the transfer characteristics from the vibration source to the environment target can be accurately obtained through the multiple regression algorithm and superposition algorithm. Compared with the traditional numerical simulation method composed of both the vehicle-track coupling model and the track-tunnel-soil model, the coefficient of determination R-squared of the two methods in the time domain can reach above 0.9, and errors of the environment evaluation index in the frequency domain are within 0.2 dB. However, the computational effort has been reduced significantly without establishing a complex numerical model of the tunnel-soil (structure) system. Moreover, it is found that environmental vibrations on the ground surface are mainly affected by fastening forces in the vicinity of 60 m centered by the observation point. In particular, the proposed method is expected to supply a certain reference for predicting vibrations under multi-source load in mechanical science and civil engineering.

Optimizing throughput of Seq2Seq model training on the IPU platform for AI-accelerated CFD simulations

Intelligence Processing Units (IPU) have proven useful for many AI applications. In this paper, we evaluate them within the emerging field of AI for simulation, where traditional numerical simulations are supported by artificial intelligence approaches. We focus specifically on a program for training machine learning models supporting a computational fluid dynamics application. We use custom TensorFlow provided by the Poplar Software Development Kit to adapt the program for the IPU-POD16 platform and investigate its ease of use and performance scalability. Training a model on data from OpenFOAM simulations allows us to get accurate simulation state predictions in test time. We describe how to optimize multi-threading runtime options and utilize the popdist library to overcome a performance bottleneck in feeding training data to the IPU on the host side. Due to communication overheads, using data parallelism to utilize two IPUs instead of one does not improve the throughput. However, once the intra-IPU costs have been paid, the hardware capabilities for inter-IPU communication allow for good scalability. Increasing the number of IPUs from two to 16 improves the throughput from 560.8 to 2805.8 samples/s. Additionally, the experimental results show that reducing the precision of input data storage from FP32 to FP16 allows to improve training throughput by 12%, while tuning selected runtime variables, by up to 6.3%.

A curved boundary treatment for discrete Boltzmann model of shallow water flows based on a partially saturated method

In the present study, a boundary treatment scheme named the partially saturated method (PSM) is incorporated into the discrete Boltzmann model (DBM) for shallow water flows to deal with complex curved boundaries. In this PSM-DBM, the two-dimensional 16 velocity levels scheme is adopted and the finite difference method is used to solve the governing equation. Then, the PSM-DBM has been applied to simulate four cases, i.e. such as steady flow in a bending channel, the flow past a stationary cylinder, a jet-forced flow in a circular basin and flow in a meandering channel with 90° consecutive bends. The simulated results have been compared with the experiments and the simulation by traditional numerical simulation. The study shows that the agreement between simulation and experiments is good. It is demonstrated that the PSM-DBM is stable and accurate to deal with the curved boundaries. Moreover, the implementation of the PSM is relatively straightforward in treating stationary curved geometries and is easy to be incorporated into the DBM for shallow water flows. In conclusion, the proposed PSM-DBM can be used widely for flows with curved boundaries.