Numerical Prediction Method Research Articles

Numerical method for dynamic prediction of temperature field in aviation gear profile grinding

Numerical method for dynamic prediction of temperature field in aviation gear profile grinding

MAFNet: Multimodal Asymmetric Fusion Network for Radar Echo Extrapolation

Radar echo extrapolation (REE) is a crucial method for convective nowcasting, and current deep learning (DL)-based methods for REE have shown significant potential in severe weather forecasting tasks. Existing DL-based REE methods use extensive historical radar data to learn the evolution patterns of echoes, they tend to suffer from low accuracy. This is because data of radar modality face difficulty adequately representing the state of weather systems. Inspired by multimodal learning and traditional numerical weather prediction (NWP) methods, we propose a Multimodal Asymmetric Fusion Network (MAFNet) for REE, which uses data from radar modality to model echo evolution, and data from satellite and ground observation modalities to model the background field of weather systems, collectively guiding echo extrapolation. In the MAFNet, we first extract overall convective features through a global shared encoder (GSE), followed by two branches of local modality encoder (LME) and local correlation encoders (LCEs) that extract convective features from radar, satellite, and ground observation modalities. We employ an multimodal asymmetric fusion module (MAFM) to fuse multimodal features at different scales and feature levels, enhancing radar echo extrapolation performance. Additionally, to address the temporal resolution differences in multimodal data, we design a time alignment module based on dynamic time warping (DTW), which aligns multimodal feature sequences temporally. Experimental results demonstrate that compared to state-of-the-art (SOTA) models, the MAFNet achieves average improvements of 1.86% in CSI and 3.18% in HSS on the MeteoNet dataset, and average improvements of 4.84% in CSI and 2.38% in HSS on the RAIN-F dataset.

Spatio-Temporal Heterogeneous Ensemble Learning Method for Predicting Land Subsidence

The prediction of land subsidence is of significant value for the early warning and prevention of geological disasters. Although numerous land subsidence prediction methods are currently available, two obstacles still exist: (i) spatio-temporal heterogeneity of land subsidence is not well considered, and (ii) the prediction performance of individual models is unsatisfactory when the data do not meet their assumptions. To address these issues, we developed a spatio-temporal heterogeneous ensemble learning method for predicting land subsidence. Firstly, a two-stage hybrid spatio-temporal clustering method was proposed to divide the dataset into internally homogeneous spatio-temporal clusters. Secondly, within each spatio-temporal cluster, an ensemble learning strategy was employed to combine one time series prediction model and three spatio-temporal prediction models to reduce the prediction uncertainty of an individual model. Experiments on a land subsidence dataset from Cangzhou, China, show that the prediction accuracy of the proposed method is significantly higher than that of four individual prediction models.

Analytical time-coordinated entry guidance for multi-hypersonic vehicles within three-dimensional corridor

A coordinated entry mission requires multiple hypersonic vehicles to arrive at the designated area simultaneously from different times and locations. To satisfy the coordinated entry mission, this paper investigates the coordinated entry guidance problem for multi-hypersonic vehicles. An analytical time-coordinated entry guidance method is proposed based on the algorithms of analytical prediction, profile correction, and coordination time determination. Firstly, considering the altitude variations of the trajectory, high-precision analytical solutions for the remaining flight time and range are derived based on quasi-equilibrium glide conditions. Compared to the traditional numerical integration prediction methods, this analytical prediction method achieves superior computational efficiency while ensuring high-precision prediction. Subsequently, a flight corridor based on longitudinal lift-to-drag ratio is constructed, extending from the traditional two-dimensional corridor to a three-dimensional one. Within the corridor, the analytical lift-to-drag ratio profile is corrected by minimizing the prediction errors, and the guidance commands for both angle of attack and bank angle can be generated directly without tracking law. With the above analytical solutions and algorithms, the coordinated entry guidance method enables rapid generation of guidance commands for each hypersonic vehicle. Finally, numerical simulations under dispersed launching time and launching location conditions are performed, and the simulation results verify the effectiveness and robustness of the proposed time-coordination entry guidance algorithm.

Numerical prediction of passive speed and performance for multistage pump without power drive in natural flow process

With the development of engineering applications and the increase in system complexity, some particular fields, such as liquid rocket engine turbopumps, aircraft engine fuel systems, and marine natural flow cooling systems, are increasingly focusing on the performance characteristics of pumps under natural flow conditions. The pump is in the form of resistance components under natural flow conditions without a power drive. The impeller undergoes passive rotation by the impact of inlet flow. Due to the specificity of its operating conditions and performance indicators, the pump's natural flow performance cannot be evaluated by regular methods. Therefore, this paper proposed a numerical prediction method for pump natural flow performance based on a coupled computational fluid dynamics coupled with six-degrees-of freedom model. The performance of a multistage pump with guide vanes was evaluated under different natural flow conditions, and the accuracy was verified by experimental measurements. The transient variation mode of pump performance parameters with time at the initial stage of natural flow impact was analyzed. The flow field's transient evolution characteristics and the wall shear stress variation during natural flow were investigated. It was found that the impeller's passive rotational speed increases linearly with the natural flow rate, while the hydraulic loss shows an exponentially increasing trend. Meanwhile, the natural flow loss coefficient shows an exponentially decreasing trend and gradually tends to a stable value. The high turbulent kinetic energy inside the impeller is mainly distributed in the flow separation region and large velocity gradients. The distribution of shear stresses is closely related to the flow behavior inside the pump and the geometrical features of the hydraulic components.

Investigation on Calm Water Resistance of Wind Turbine Installation Vessels with a Type of T-BOW

Given the typical characteristics of self-propulsion and jack-up wind turbine installation vessels (WTIVs), including their full and blunt hull form and complex appendages, this paper combines the model test method with the RANS-based CFD numerical prediction method to experimentally and numerically study the resistance of the optimized hull at different spudcan retraction positions. The calm water resistance components and their mechanisms of WTIVs based on T-BOW were obtained. Furthermore, using the multivariate nonlinear least squares method, an empirical formula for rapid resistance estimation based on the Holtrop method was derived, and its prediction accuracy and applicability were validated with a full-scale ship case. This study indicates that the primary resistance components of such low-speed vessels are viscous pressure resistance, followed by frictional resistance and wave-making resistance. Notably, the spudcan retraction well area, as a unique appendage of WTIVs, exhibits a significant “moonpool additional resistance” effect. Different spudcan retraction positions affect the total calm water resistance by approximately 20% to 30%. Therefore, in the resistance optimization design of WTIVs, special attention should be paid to the matching design of the spudcan structure and the hull shell plate lines in the spudcan retraction well area.

A miRNA-disease association prediction model based on tree-path global feature extraction and fully connected artificial neural network with multi-head self-attention mechanism

BackgroundMicroRNAs (miRNAs) emerge in various organisms, ranging from viruses to humans, and play crucial regulatory roles within cells, participating in a variety of biological processes. In numerous prediction methods for miRNA-disease associations, the issue of over-dependence on both similarity measurement data and the association matrix still hasn’t been improved. In this paper, a miRNA-Disease association prediction model (called TP-MDA) based on tree path global feature extraction and fully connected artificial neural network (FANN) with multi-head self-attention mechanism is proposed. The TP-MDA model utilizes an association tree structure to represent the data relationships, multi-head self-attention mechanism for extracting feature vectors, and fully connected artificial neural network with 5-fold cross-validation for model training.ResultsThe experimental results indicate that the TP-MDA model outperforms the other comparative models, AUC is 0.9714. In the case studies of miRNAs associated with colorectal cancer and lung cancer, among the top 15 miRNAs predicted by the model, 12 in colorectal cancer and 15 in lung cancer were validated respectively, the accuracy is as high as 0.9227.ConclusionsThe model proposed in this paper can accurately predict the miRNA-disease association, and can serve as a valuable reference for data mining and association prediction in the fields of life sciences, biology, and disease genetics, among others.Graphical

Experimental and Numerical Investigation Integrated with Machine Learning (ML) for the Prediction Strategy of DP590/CFRP Composite Laminates.

This study unveils a machine learning (ML)-assisted framework designed to optimize the stacking sequence and orientation of carbon fiber-reinforced polymer (CFRP)/metal composite laminates, aiming to enhance their mechanical properties under quasi-static loading conditions. This work pioneers the expansion of initial datasets for ML analysis in the field by uniquely integrating the experimental results with finite element simulations. Nine ML models, including XGBoost and gradient boosting, were assessed for their precision in predicting tensile and bending strengths. The findings reveal that the XGBoost and gradient boosting models excel in tensile strength prediction due to their low error rates and high interpretability. In contrast, the decision trees, K-nearest neighbors (KNN), and random forest models show the highest accuracy in bending strength predictions. Tree-based models demonstrated exceptional performance across various metrics, notably for CFRP/DP590 laminates. Additionally, this study investigates the impact of layup sequences on mechanical properties, employing an innovative combination of ML, numerical, and experimental approaches. The novelty of this study lies in the first-time application of these ML models to the performance optimization of CFRP/metal composites and in providing a novel perspective through the comprehensive integration of experimental, numerical, and ML methods for composite material design and performance prediction.

A Survey of Deep Learning for Detecting miRNA- Disease Associations: Databases, Computational Methods, Challenges, and Future Directions.

MicroRNAs (miRNAs) are an important class of non-coding RNAs that play an essential role in the occurrence and development of various diseases. Identifying the potential miRNA-disease associations (MDAs) can be beneficial in understanding disease pathogenesis. Traditional laboratory experiments are expensive and time-consuming. Computational models have enabled systematic large-scale prediction of potential MDAs, greatly improving the research efficiency. With recent advances in deep learning, it has become an attractive and powerful technique for uncovering novel MDAs. Consequently, numerous MDA prediction methods based on deep learning have emerged. In this review, we first summarize publicly available databases related to miRNAs and diseases for MDA prediction. Next, we outline commonly used miRNA and disease similarity calculation and integration methods. Then, we comprehensively review the 48 existing deep learning-based MDA computation methods, categorizing them into classical deep learning and graph neural network-based techniques. Subsequently, we investigate the evaluation methods and metrics that are frequently used to assess MDA prediction performance. Finally, we discuss the performance trends of different computational methods, point out some problems in current research, and propose 9 potential future research directions. Data resources and recent advances in MDA prediction methods are summarized in the GitHub repository https://github.com/sheng-n/DL-miRNA-disease-association-methods.

Surrogate numerical prediction method of TBM position via FEM simulation and machine learning

The position of the tunnel boring machine (TBM) during tunnel construction is critical and must be precisely controlled. Owing to the geological uncertainty and complexity of the interaction between the ground and the TBM, controlling the position of the TBM is challenging. Hence, a surrogate numerical method is proposed to predict the position of the TBM using the finite element method (FEM) and machine learning method. First, a refined three-dimensional FEM model was established. Different values of the property parameters of the ground and the thrust force of the TBM were input into the FEM model, generating a database that includes 1000 cases. Subsequently, the database is used to train a gradient boosting regression (GBR) model. The GBR model learns from the database and establishes the relationship between the construction parameters, ground parameters, and TBM position as a surrogate model. The surrogate model exhibited high accuracy on the test set. With a geological survey and a construction parameter monitoring system, the TBM position could be predicted quickly and precisely using a surrogate model. The construction parameters were adjusted if the TBM position did not satisfy the requirements.

Numerical generation and verification of a two-dimensional virtual asphalt mixture

This paper aims at finding a robust method to generate a two-dimensional virtual asphalt mixture and to verify the suitability of the generated virtual mixture. First, three methods namely stereological method, Walraven formula and phase separation method are adopted to convert the actual three-dimensional gradation curves into two-dimensional ones. Then, an approach called semi whole scale cross section (SWSCS) method is developed to verify the results obtained and to determine the most robust procedure to follow. Third, the numerical two-dimensional geometry shape data library (GSDL) of different aggregates and voids and a robust method to assemble representative volume elements (RVEs) are developed using Matlab. Finally, a gradation curve from the literature is considered to generate a virtual asphalt mixture specimen, which is then imported in the finite element (FE) software Abaqus to simulate the dynamic modulus test. The virtual test results are further verified using available literature and another homogenization numerical prediction method based on the Mori-Tanaka's approach. The results show that the phase separation method is effective to convert the three-dimensional gradation curve to a two-dimensional one. The simulation results of the virtual asphalt mixture developed in this study turn out to be in good agreement with both previously presented literature data and Mori-Tanaka's theory.

Identifying causes of errors between two wave-related data using performance metrics

Recently, numerous prediction methods of time-series data of wave-related phenomena have been developed. Quantitative evaluation of an error of a predicted result against a reference result is important to improve prediction accuracy. Many performance metrics are engaged to evaluate their accuracy. However, it is difficult for them to identify the error causes because all errors, no matter what the cause, are combined together. This paper presents new representations of performance metrics to separate errors by causes. Performances are evaluated in a frequency domain instead of a time domain. A frequency domain’s amplitude and phase are respectively evaluated using performance metrics. In addition, to detect errors due to instantaneous phenomena and changes in an error trend over time, mean errors are defined by three-time intervals. The original metric uses all time-series data. On the other hand, a finite interval mean error at the present time is calculated by the mean of the preceding data. In addition, cumulative mean error at the present time is calculated by the mean of the data up to the present time. These new representations of performance metrics could help to identify error causes. Benchmark tests are carried out to demonstrate the validity of the proposed representations.

Using Automated Machine Learning for Spatial Prediction—The Heshan Soil Subgroups Case Study

Recently, numerous spatial prediction methods with diverse characteristics have been developed. Selecting an appropriate spatial prediction method, along with its data preprocessing and parameter settings, presents a challenging task for many users, especially for non-experts. This paper addresses this challenge by exploring the potential of automated machine learning method proposed in artificial intelligent domain to automatically determine the most suitable method among various machine learning methods. As a case study, the automated machine learning method was applied to predict the spatial distribution of soil subgroups in Heshan farm. A total of 110 soil samples and 10 terrain variables were utilized in the designed experiments. To evaluate the performance, the proposed method was compared to each machine learning method with default parameters values or parameters determined by expert knowledge. The results showed that the proposed method typically achieved higher accuracy scores than the two alternative methods. This suggests that automated machine learning performs effectively in scenarios where numerous machine learning methods are available and offers practical utility in reducing the dependence on users’ expertise in spatial prediction. However, a more robust automated framework should be developed to encompass a broader range of spatial prediction methods, such as spatial statistic methods, rather than only focusing on machine learning methods.

Study of 3D-printed onyx parts reinforced with continuous glass fibers: Focus on mechanical characterization, analytical prediction and numerical simulation

The 3D printing of continuous-fiber composites is currently relevant to engineers and researchers. This study aims to characterize and predict the mechanical properties of Onyx/glass fiber specimens printed using 3D printing. The work assesses the impact of glass fiber printing parameters on the mechanical behavior of printed parts and proposes analytical and numerical methods to predict mechanical properties. A physicochemical analysis was conducted on 3D printed continuous glass fibers. The study also investigated the impact of fiber printing parameters on composite parts. The results indicate that the 3D-printed glass fibers consist of nylon, continuous glass fibers, and voids (porosity), which range from 58% to 63%, 31% to 38%, and 5% to 8%, respectively. Mechanical characterizations indicate that printing fiber layers in blocks results in superior mechanical properties compared to printing alternating layers of glass fibers and Onyx. Additionally, the concentric mode of fiber printing can be challenging if the ‘start rotation’ parameter is not adjusted correctly. Premature specimen breakage occurred when fiber printing began within their useful length, resulting in a deformation at break that was approximately 34% less, depending on the starting position. The proposed analytical and numerical prediction methods had prediction errors of approximately 7% to 12% and 5% to 7%, respectively. Engineers can use these prediction approaches during the dimensioning phase of 3D printed composite parts.

Link-Backdoor: Backdoor Attack on Link Prediction via Node Injection

Link prediction, inferring the undiscovered or potential links of the graph, is widely applied in the real world. By facilitating labeled links of the graph as the training data, numerous deep learning-based link prediction methods have been studied, which have dominant prediction accuracy compared with nondeep methods. However, the threats of maliciously crafted training graphs will leave a specific backdoor in the deep model; thus, when some specific examples are fed into the model, it will make a wrong prediction, defined as a backdoor attack. It is an important aspect that has been overlooked in the current literature. In this article, we prompt the concept of a backdoor attack on link prediction and propose Link-Backdoor to reveal the training vulnerability of the existing link prediction methods. Specifically, the Link-Backdoor combines the fake nodes with the nodes of the target link to form a trigger. Moreover, it optimizes the trigger by the gradient information from the target model. Consequently, the link prediction model trained on the backdoored dataset will predict the link with a trigger to the target state. Extensive experiments on five benchmark datasets and five well-performing link prediction models demonstrate that the Link-Backdoor achieves the state-of-the-art attack success rate under both the white box (i.e., available of the target model parameter) and the black box (i.e., unavailable of the target model parameter) scenarios. In addition, we testify the attack under the defensive circumstance, and the results indicate that the Link-Backdoor still can construct a successful attack on the well-performing link prediction methods. The code and data are available at https://github.com/Seaocn/Link-Backdoor.

Respiratory Motion Prediction with Empirical Mode Decomposition-Based Random Vector Functional Link

The precise prediction of tumor motion for radiotherapy has proven challenging due to the non-stationary nature of respiration-induced motion, frequently accompanied by unpredictable irregularities. Despite the availability of numerous prediction methods for respiratory motion prediction, the prediction errors they generate often suffer from large prediction horizons, intra-trace variabilities, and irregularities. To overcome these challenges, we have employed a hybrid method, which combines empirical mode decomposition (EMD) and random vector functional link (RVFL), referred to as EMD-RVFL. In the initial stage, EMD is used to decompose respiratory motion into interpretable intrinsic mode functions (IMFs) and residue. Subsequently, the RVFL network is trained for each obtained IMF and residue. Finally, the prediction results of all the IMFs and residue are summed up to obtain the final predicted output. We validated this proposed method on the benchmark datasets of 304 respiratory motion traces obtained from 31 patients for various prediction lengths, which are equivalent to the latencies of radiotherapy systems. In direct comparison with existing prediction techniques, our hybrid architecture consistently delivers a robust and highly accurate prediction performance. This proof-of-concept study indicates that the proposed approach is feasible and has the potential to improve the accuracy and effectiveness of radiotherapy treatment.

Numerical Study of the Effect of Primary Nozzle Geometry on Supersonic Gas-Solid Jet of Bypass Injected Dry Powder Fire Extinguishing Device

A two-way coupled model between polydisperse particle phases with compressible gases and a density-based coupling implicit solution method, combining the third-order MUSCL with QUICK spatial discretization scheme and the second-order temporal discretization scheme, are constructed based on the discrete-phase model (DPM) and the stochastic wander model (DRWM) in the Eulerian–Lagrangian framework in conjunction with a unitary particulate source (PSIC) approach and the SST k-ω turbulence model. The accuracy of the numerical prediction method is verified using previous supersonic nozzle gas-solid two-phase flow experiments. Numerical simulation of a two-phase jet of dry powder extinguishing agent gas with pilot-type supersonic nozzle was performed to analyze the influence of geometrical parameters, such as the length ratio rL and the area ratio rA of the main nozzle on the two-phase flow field, as well as on the jet performance indexes, such as the particle mean velocity vp,a, velocity inhomogeneity Φvp, particle dispersion Ψp, particle mean acceleration ap,a, etc. By analyzing the parameters, we indicate the requirements for the combination of jet performance metrics for different flame types such as penetrating, spreading, and dispersing.

Cracking the black box of deep sequence-based protein-protein interaction prediction.

Identifying protein-protein interactions (PPIs) is crucial for deciphering biological pathways. Numerous prediction methods have been developed as cheap alternatives to biological experiments, reporting surprisingly high accuracy estimates. We systematically investigated how much reproducible deep learning models depend on data leakage, sequence similarities and node degree information, and compared them with basic machine learning models. We found that overlaps between training and test sets resulting from random splitting lead to strongly overestimated performances. In this setting, models learn solely from sequence similarities and node degrees. When data leakage is avoided by minimizing sequence similarities between training and test set, performances become random. Moreover, baseline models directly leveraging sequence similarity and network topology show good performances at a fraction of the computational cost. Thus, we advocate that any improvements should be reported relative to baseline methods in the future. Our findings suggest that predicting PPIs remains an unsolved task for proteins showing little sequence similarity to previously studied proteins, highlighting that further experimental research into the 'dark' protein interactome and better computational methods are needed.

DeepWind: a heterogeneous spatio-temporal model for wind forecasting

Deep learning (DL) has shown great potential in enhancing the performance of traditional numerical weather prediction (NWP) methods in weather forecasting. Certain applications such as wind power generation desire more accurate wind predictions which is challenging due to limited observations and complex dynamics. To this end, this paper introduces a DL-based heterogeneous network named DeepWind for NWP correction, which can simultaneously correct the NWP of diverse wind variables across multiple weather stations. In particular, it first exerts the meteorological domain knowledge to achieve an effective transformation of target variables and then develops a heterogeneous neural network to learn spatio-temporal representations. A novel difference loss function is further designed for stable temporal learning. Moreover, this study might be the first to expose an underlying evaluation problem in deep forecasting, which we call evaluation inconsistency, thereby necessitating the assessment of model performance across diverse evaluation metrics. Experimental results demonstrate the superiority of the proposed approach over strong DL baselines, which makes it positioned for deployment in the real-world production environment. Source code is released at https://github.com/Rittersss/DeepWind.

Numerical Method to Determine the Inception of Propeller Tip Vortex Cavitation Based on Bubble Dynamics

Due to the scale effect, tip vortex cavitation (TVC) is the earliest type of cavitation that occurs on real ship propellers. As a result, experts in the ship field have been paying close attention to the accurate prediction of propeller TVC inception for a long time. The motion and growth of the microscopic nuclei in the water have a significant influence on TVC inception. However, the minimum pressure coefficient method—a common method at present—based on the traditional Eulerian framework, neglects the influence of microscopic nuclei and therefore cannot accurately predict the cavitation inception. Moreover, the numerical prediction method for cavitation inception, which is based on bubble dynamics models and considers the influence of nuclei, has not established a set of unified and specific discrimination criteria applicable to propeller cavitation inception. In order to make up for the shortcomings of traditional prediction models and the existing methods based on bubble dynamics in the prediction of TVC inception, we propose a new discrimination method for propeller TVC inception based on bubble dynamics in this paper. The comparison with experimental results demonstrates that our proposed method allows us to predict propeller TVC inception more accurately. In addition, the effect mechanism of tip vortex flow characteristics on nuclei evolution is further investigated, and it is found that when approaching the low-pressure region at a vortex core under the influence of tip vortex suction, nuclei grow explosively under the continuous action of the low pressure at the vortex core until they reach their maximum sizes and then collapse rapidly.