Computational Accuracy Research Articles

Upwind block-centered multistep differences for semiconductor device problem with heat and magnetic influences

In this paper, a mathematical model of semiconductor device with two important factors is discussed. Heat and magnetic influences are considered together. Its numerical approximation is important in information science and other technological innovation fields. The major physical unknowns (the potential, electron concentration, hole concentration, and the heat) are defined by four nonlinear PDEs: an elliptic equation, two convection-diffusion equations and a heat conductor equation. A conservative block-centered method is used to approximate the elliptic equation. The derivatives to time t, ∂∂t(⋅), are discretized by multistep differences for improving the computational accuracy. The diffusions and convection terms are treated by block-centered differences and upwind approximations, respectively. This composite numerical method is suitable for this nonlinear system. Firstly, numerical dispersion and nonphysical oscillations are avoided. Secondly, the unknowns and their adjoint vectors are obtained simultaneously. Thirdly, the law of conservation is preserved. An error estimates is derived. Finally, simple examples are given for showing the efficiency and accuracy.

Seismic performance of mortise-tenon joint in prefabricated platform canopy

To optimize the prefabricated canopy joints, improve the construction efficiency, and enhance the seismic performance of assembly joints, a novel mortise-tenon joint (MTJ) was proposed. To evaluate the seismic performance along the vertical and track directions, two 1:1.5 scaled specimens were tested under low-cycle reciprocated loading. The test results find the hysteretic curves in both directions exhibit the “pinch” effect especially in the vertical direction, indicating the feature of shear slippage. The displacement ductility coefficients were all above 5.0, and the ultimate drift ratios were 4.54% and 5.88% respectively, showing excellent deformation ability and ductility. The parameters, such as the sectional strength ratio, the stirrup ratio of the prefabricated beam, the reinforcement ratio and the cross-sectional area of the enlarged section, were analyzed by the validated finite element (FE) models. The analysis shows as the section strength ratio decreases and the stirrup ratio increased, the seismic bearing capacity improved, while the plastic hinge located on the prefabricated column varied to the prefabricated beam ends. Based on the results, it is recommended the number of longitudinal rebars through the prefabricated beam is 4, the stirrup diameter in the prefabricated beam is 10mm and the the column cap is no less than 400mm. Finally, based on the truss-diagonal strut model, a method for calculating the shear strength capacity of the novel joint was proposed. Compared with the test and FE results, the theoretical equations exhibit good computational accuracy.

Determining the Optimal Tradeoff between Compute and Accuracy for Diffusion Models in Molecular Docking

In the world of molecular biology, Molecular Docking has become a pillar for understanding complex interactions pivotal for drug design and more. The emergence of sophisticated machine learning models, such as Diffusion Generative Models, have significantly expanded our analytical capabilities. However, this evolution has introduced notable computational challenges. This study aims to examine the trade-off between computational demands and model accuracy. Our methodology, which incorporates free platforms, illuminates methods to conserve computational resources while maintaining nearoptimal accuracy. Our findings suggest that using 30 samples per complex, 15 inference steps, and 4 batch steps improves pose prediction accuracy and reduces computational resources. The proposed parameters achieve a 14% accuracy increase compared to the 40 samples per complex model and a 56.25% increase compared to the 10 samples per complex model. The optimized inference steps result in a 12.2% accuracy increase over the 20-step control using the 40 samples model and a 24.3% increase using the 10 samples model. Additionally, using 4 batch steps leads to a 40.6% increase in DiffDock Confidence for the 10-sample control and a 0.4% increase for the 40- sample control.

MatSwarm: trusted swarm transfer learning driven materials computation for secure big data sharing

The rapid advancement of Industry 4.0 necessitates close collaboration among material research institutions to accelerate the development of novel materials. However, multi-institutional cooperation faces significant challenges in protecting sensitive data, leading to data silos. Additionally, the heterogeneous and non-independent and identically distributed (non-i.i.d.) nature of material data hinders model accuracy and generalization in collaborative computing. In this paper, we introduce the MatSwarm framework, built on swarm learning, which integrates federated learning with blockchain technology. MatSwarm features two key innovations: a swarm transfer learning method with a regularization term to enhance the alignment of local model parameters, and the use of Trusted Execution Environments (TEE) with Intel SGX for heightened security. These advancements significantly enhance accuracy, generalization, and ensure data confidentiality throughout the model training and aggregation processes. Implemented within the National Material Data Management and Services (NMDMS) platform, MatSwarm has successfully aggregated over 14 million material data entries from more than thirty research institutions across China. The framework has demonstrated superior accuracy and generalization compared to models trained independently by individual institutions.

GBNSS: A Method Based on Graph Neural Networks (GNNs) for Global Biological Network Similarity Search

Biological network similarity search plays a crucial role in the analysis of biological networks for human disease research and drug discovery. A biological network similarity search aims to efficiently identify novel networks biologically homologous to the query networks. Great progress has been achieved in biological network similarity searches. However, it remains a challenge to mine the biological network information fully to improve the accuracy of query results without increasing time overheads. In this study, we propose a biological network similarity search method based on graph neural networks named GBNSS, which combines topological and biological information (GO annotations) of biological networks into graph neural networks to find topologically and biologically similar biological networks in the database. Additionally, GBNSS is a topology-free biological network similarity search method with an arbitrary network structure. The experimental results on four benchmark datasets show that GBNSS outperforms the existing methods in terms of computational efficiency and search accuracy. Case studies further demonstrate that GBNSS is capable of searching similar networks in real-world biological networks.

Detection of optic disc in human retinal images utilizing the Bitterling Fish Optimization (BFO) algorithm

Early detection and correct identification of the optic disc (OD) on scanned retinal images are significant for diagnosing and treating several ophthalmic conditions, including glaucoma and diabetic retinopathy. Conventional methods for detecting the OD often struggle with processing retinal images due to noise, changes in illumination, and complex overlapping images. This study presents the development of effective and accurate fixation of the optic disc using the Bitterling Fish Optimization (BFO) algorithm, which enhances the processes of OD imaging in speed and precision. The proposed method begins with image enhancement and noise suppression for preprocessing, followed by applying the BFO algorithm to locate and delineate the OD region. The performance evaluation of the algorithm was conducted within several public domain retinal images, including DRIVE, STARE, ORIGA, DRISHTI-GS, DiaRetDB0, and DiaRetDB1 datasets about some internal metrics: sensitivity (SE), specificity (SP), accuracy (ACC), DICE overlap coefficient, overlap and time of processing respectively. The technique based on BFO provided better results, with 99.33%, 99.94%, and 98.22% accuracy achieved for OD in DRIVE, DRISHTI-GS, and DiaRetDB 1, respectively. The approach also demonstrated high overlaps and good DICE results, with a DICE coefficient of 0.9501 for the DRISHTI-GS database. On average, the processing time per image was less than 2.5 s, proving the approach’s efficiency in computations. The BFO approach has demonstrated its effectiveness and scalability in detecting optic discs in retinal images in an automated manner. It showed impressive performance levels in terms of computation time and accuracy and was variation resistant irrespective of the quality of the image and the pathology present on it. This method holds significant potential for clinical use, providing a meaningful way of diagnosing and managing ocular disease at an early stage.

On the Numerical Investigation of Natural-Convection Heat Sinks Across a Wide Range of Flow and Operating Conditions

Many designs for natural-convection heat sinks and semi-empirical correlations have been proposed in the recent years, but they are only valid in a limited range of Elenbaas numbers El and were mostly tested for laminar flows. To alleviate those limits, parametric studies with 2D and quasi-3D models were carried out, in ranges of Grashof numbers up to 1.55×1011 and Elenbaas numbers up to 3.42×107. Ansys Fluent’s laminar, transition-SST, SST k-ω and k-ϵ models were applied. In addition, when used in this valid range, i.e., mean Elenbaas numbers, with the simplified quasi-3D model, the transition-SST model could predict better results, overestimating the heat flux by 10 to 15% compared to semi-empirical correlations. The 2D model was not deemed satisfying, regarding turbulence models. Consequently, a quasi-3D model was developed: it appeared to be an efficient trade-off between computational time and prediction accuracy, in particular for turbulence models. New grouping factors were also found, to ensure proper dimensioning of natural-convection heat sinks. They corresponded to non-dimensional parameters that dictated the physical behaviour of the heat sink with respect to the semi-empirical correlations. Typically, the ratio of the spacing to the optimal spacing predicted by Bar-Cohen’s correlation turned out to be an appropriate grouping factor with a threshold of 1, above which the fins could safely be considered as isolated, thus greatly simplifying all further calculations.

Mechanical Analysis and Optimization of Concrete Structures Based on Advanced Finite Element Method

This article explores the mechanical analysis and optimization problems of concrete structures based on the Advanced Finite Element Method. By integrating advanced numerical techniques with practical engineering cases, this study aims to improve the safety and economy of concrete structure design. Firstly, the paper outlines the limitations of traditional finite element methods in concrete structure analysis, such as insufficient computational accuracy and low computational efficiency. Subsequently, by introducing AFEM, we significantly improved the accuracy and efficiency of the analysis. For example, when simulating a complex bridge structure, AFEM not only reduces the calculation time by about 25%, but also improves the accuracy of stress distribution prediction by more than 10%. In the optimization stage, we utilized the analysis results of AFEM and optimized the material consumption, cross-sectional dimensions, and reinforcement parameters of concrete structures through multi-objective optimization algorithms. A comprehensive data analysis underscores that the optimized concrete structure triumphantly meets all safety performance criteria while achieving a remarkable 12% reduction in material usage. This substantial material savings translates into a substantial 8% decrease in overall construction costs, significantly bolstering the project’s economic feasibility. Moreover, these cost savings not only amplify the project’s profitability but also play a pivotal role in enhancing the structure’s longevity and durability, thereby contributing to its sustainable performance over its entire service life. Furthermore, we delved into the capability of AFEM in simulating intricate phenomena such as the nonlinear behavior of concrete materials, crack propagation patterns, and the intricate interactions between steel reinforcements and concrete. These complex mechanical behaviors are crucial for the safety and stability of structures, and AFEM provides more comprehensive and accurate references for structural design by accurately simulating these behaviors. This article conducts in-depth research on the mechanical analysis and optimization of concrete structures based on AFEM, and demonstrates the significant advantages of AFEM in improving structural safety and economy through specific data. These research results not only provide new theoretical support and practical tools for concrete structure design, but also provide valuable references for future research and application in related fields.

A Low Mach Preconditioned Harmonic Balance Solver for Cavity Flutter Computations

Abstract Labyrinth seal flutter is a critical phenomenon in turbomachinery, as it can lead to severe structural vibrations and potential component damage. Accurate prediction and mitigation of flutter are paramount to ensuring the reliability and performance of modern turbomachinery systems. This paper explores the numerical computation of a labyrinth seal flutter test case using a low Mach preconditioned harmonic balance (HB) solver and investigates how this approach can improve the accuracy and response time of flutter computations. HB solvers have gained prominence in turbomachinery computations for their ability to efficiently capture unsteady flow phenomena and significantly reduce computational time compared to time-domain analyses. In labyrinth seals, however, the flow is often characterized by low Mach numbers, and preconditioning for these conditions has been shown to significantly improve convergence and accuracy. The goal of this paper is to demonstrate how to implement low Mach preconditioning in a HB solver in the frequency domain. We employ iterative preconditioning to alleviate the stiffness associated with density-based solvers under low Mach conditions and analyze the effect of the preconditioning parameters on the convergence rate. Furthermore, we address inaccuracies linked to the classical Roe solver in low Mach scenarios by adapting it to the low Mach preconditioned governing equations. Through the combined utilization of iterative preconditioning and a preconditioned Roe solver, this study aims to improve convergence rates and the overall quality of flutter predictions. We demonstrate the method with an academic labyrinth seal test case originally presented by Corral et al. (“Higher Order Conceptual Model for Labyrinth Seal Flutter,” ASME J. Turbomach., 143(7), p. 071006). While previous investigations have primarily relied on linearized frequency domain solvers and reduce-order models, in this research a preconditioned HB solver is applied to this test case.

Neural fields for rapid aircraft aerodynamics simulations

This paper presents a methodology to learn surrogate models of steady state fluid dynamics simulations on meshed domains, based on Implicit Neural Representations (INRs). The proposed models can be applied directly to unstructured domains for different flow conditions, handle non-parametric 3D geometric variations, and generalize to unseen shapes at test time. The coordinate-based formulation naturally leads to robustness with respect to discretization, allowing an excellent trade-off between computational cost (memory footprint and training time) and accuracy. The method is demonstrated on two industrially relevant applications: a RANS dataset of the two-dimensional compressible flow over a transonic airfoil and a dataset of the surface pressure distribution over 3D wings, including shape, inflow condition, and control surface deflection variations. On the considered test cases, our approach achieves a more than three times lower test error and significantly improves generalization error on unseen geometries compared to state-of-the-art Graph Neural Network architectures. Remarkably, the method can drastically accelerate high fidelity numerical solver, achieving a five order of magnitude speedup on the RANS transonic airfoil dataset.

A Task-driven Adversarial Channel Selection Method for Binary Classification Based on Magnetocardiography.

As the number of sensors in magnetocardiography (MCG) arrays increases to capture detailed cardiac activity, some channels contribute minimally to task performance, resulting in data redundancy and resource consumption. Although existing methods can reduce the number of channels required to meet task demands, they often struggle to balance computational time and the accuracy of the selected channels and overlook the scalability of the selected channels. This limitation means that when environmental conditions change, or when sensors malfunction, redesigning channel configurations becomes necessary, which increases experimental uncertainties. This study introduces a task-driven adversarial channel selection method tailored for binary classification of MCG signals. The optimal channel combination is determined through a group-wise search using a heuristic algorithm, whose objective function is designed to maximize the difference between the classification accuracy and cosine similarity of the selected channel. In evaluations using an MCG dataset from Qilu Hospital of Shandong University, the proposed method successfully reduced the number of channels from 36 to 5 without compromising classification performance. Furthermore, it outperforms existing hybrid sequential forward search method by achieving comparable accuracy with fewer channels, while also demonstrating superior scalability compared to both hybrid sequential forward search and pearson-rank methods. This approach strikes a balance between computational consumption and accuracy, while improving the scalability of the selected channel combinations, enhancing the efficiency and practicality of the MCG system.

Linearly Implicit Conservative Schemes for the Nonlocal Schrödinger Equation

This paper introduces two high-accuracy linearly implicit conservative schemes for solving the nonlocal Schrödinger equation, employing the extrapolation technique. These schemes are based on the generalized scalar auxiliary variable approach and the symplectic Runge–Kutta method. By integrating these advanced methods, the proposed schemes aim to significantly enhance computational accuracy and efficiency, while maintaining the essential conservative properties necessary for accurate physical modeling. This offers a structured approach to handle auxiliary variables, ensuring stability and conservation, while the symplectic Runge–Kutta method provides a robust framework with high accuracy. Together, these techniques offer a powerful and reliable approach for researchers dealing with complex quantum mechanical systems described by the nonlocal Schrödinger equation, ensuring both accuracy and stability in their numerical simulations.

Levee Soil Stratification Based on PAM Cluster Analysis of Measured Soil Samples from Multiple Probe Drilling Sites

Accurate soil stratification is crucial for levee safety evaluation, yet limited field sampling often hinders comprehensive analysis. This study applies the Partitioning Around Medoids (PAM, also known as K-Medoids) clustering approach for levee soil stratification using data from multiple probe drilling sites. Focusing on a Yellow River levee section in China as a study case, the PAM clustering approach effectively identifies its distinct soil types and reconstructs its soil stratification by analyzing key soil properties relevant to levee seepage and stability characteristics, including coefficient of permeability, angle of internal friction, and cohesion. The resulting soil stratification, when applied to seepage and stability analyses of the levee section, yields relatively high safety factors, indicating low failure risks under design flood conditions. These analytical results align with recent monitoring records, validating the effectiveness of the approach. A sensitivity analysis on the number of clusters, the key parameter in the PAM clustering approach, demonstrates the typical existence of an optimal value balancing computational accuracy and practical interpretability. A comparison with a hierarchical clustering approach further confirms the robustness of the PAM clustering approach. This study contributes to improving levee soil stratification methodology and enhancing levee safety evaluation, particularly when dealing with limited and spatially distributed sampling data.

Three-dimensional finite element analysis of biological signal feedback in the mechanical properties of sound production

In response to the problems of low reliability and long time consumption in traditional research on sound production, this article used electromyography (EMG) signals as input signals to build a high-precision sound production mechanics model. A Proportional-Integral-Derivative (PID) controller was used to dynamically adjust the model and develop a real-time feedback system. This article established a detailed three-dimensional (3D) finite element model including the vocal cords and throat, defined the nonlinear elastic properties and linear elastic properties of different tissues, and used tetrahedral grid partitioning technology to improve the computational accuracy of the model. Through a EMG sensor, an individual EMG was collected and filtered to remove noise. The processed EMGs were used as input parameters for the finite element model to drive the muscle units in the model. By using a PID controller to receive real-time EMG input, the error was calculated and the model was adjusted, enabling accurate simulation of the mechanical properties of vocal cord vibration under different vocal states and achieving real-time feedback. Considering the complexity of vocal cord vibration driven by biological signals, this article simulated and analyzed the modal characteristics of vocal cord vibration, and analyzed the differences in vocal cord vibration characteristics under different vocal states. The sound pressure distribution and resonance frequency were simulated and analyzed to understand the propagation characteristics of sound. Finally, by comparing and analyzing the simulated data with the actual collected data, it was found that the maximum relative error rate of the model under different sound states was 6.14%, and the overall error rate of the model was relatively low, which verified the reliability of the model. The findings demonstrated that the feedback delays of the model in different sound states were all within 100 milliseconds, indicating that the system had high real-time performance and accuracy, which was promising in practical applications.

Transfer Learning-Based Physics-Informed Convolutional Neural Network for Simulating Flow in Porous Media with Time-Varying Controls

A physics-informed convolutional neural network (PICNN) is proposed to simulate two-phase flow in porous media with time-varying well controls. While most PICNNs in the existing literature worked on parameter-to-state mapping, our proposed network parameterizes the solutions with time-varying controls to establish a control-to-state regression. Firstly, a finite volume scheme is adopted to discretize flow equations and formulate a loss function that respects mass conservation laws. Neumann boundary conditions are seamlessly incorporated into the semi-discretized equations so no additional loss term is needed. The network architecture comprises two parallel U-Net structures, with network inputs being well controls and outputs being the system states (e.g., oil pressure and water saturation). To capture the time-dependent relationship between inputs and outputs, the network is well designed to mimic discretized state-space equations. We train the network progressively for every time step, enabling it to simultaneously predict oil pressure and water saturation at each timestep. After training the network for one timestep, we leverage transfer learning techniques to expedite the training process for a subsequent time step. The proposed model is used to simulate oil–water porous flow scenarios with varying reservoir model dimensionality, and aspects including computation efficiency and accuracy are compared against corresponding numerical approaches. The comparison with numerical methods demonstrates that a PICNN is highly efficient yet preserves decent accuracy.

Open shop scheduling with group and transportation operations by learning-driven hyper-heuristic algorithms

Open shop scheduling problems (OSSPs) are complex scheduling problems, which have been extensively studied in the literature. Group and transportation activities are two important aspects of OSSPs that still need attention. This work considers an OSSP with group and transportation operations to minimize maximum completion time by solving three key sub-problems: job assignment among groups, job sequence in groups and group sequence on machines. Firstly, an integer programming model is formulized to define the problem. Secondly, a learning-driven hyper-heuristic algorithm is developed by incorporating a Q-learning method and four meta-heuristics, i.e., genetic algorithm, artificial bee colony optimization, variable neighborhood search method and Jaya algorithm. The Q-learning method is devised to select the most promising meta-heuristic for performing at each iteration. Three neighborhood structures are designed by integrating critical machines and critical paths. Finally, the developed model is verified by an exact solver CPLEX, and the comparison results exhibit that CPLEX is effective for instances with ten jobs. For the instances with more than ten jobs, the developed algorithm wins CPLEX in terms of computation accuracy and efficiency, signifying its excellent performance in finding better solutions. Furthermore, four meta-heuristics mentioned above and three state-of-the-art meta-heuristics are employed for comparisons in solving a set of benchmark test instances. The results confirm that the formulated model and algorithm have stronger competitiveness in handling the considered problems.

Position and orientation estimation method based on 3D digital morphology contour registration

Abstract Accurately and quickly obtaining the positions and orientations of mechanical parts based on the digital morphologies of mechanical parts are the key to achieving efficient and accurate assembly of mechanical parts. However, due to poor robustness and compactness in extracting digital morphology contours of mechanical parts, the accuracy of assembly positions and orientations obtained by using digital morphology contours cannot meet the requirements of high-precision assembly. Therefore, this paper proposes a position and orientation estimation method based on 3D digital morphology contour registration. This method extracts and optimizes digital morphology contours of mechanical parts and obtains the assembly positions and orientations by using an improved iterative closest point method to register the extracted digital morphology contours with those of the mechanical parts in assembly targets with desired positions and orientations. Experiments are conducted using mechanical parts from the ABC dataset and inertial confinement fusion micro-target. From the experimental results, when using the assembly positions and orientations obtained through the method proposed in this paper to assemble mechanical parts, it can achieve translation absolute errors of 8 μm, 5 μm, and 9 μm along the X-, Y-, and Z-axes, respectively. Similarly, the angular absolute errors in rotations around the Z-, Y-, and X-axes can be less than or equal to 0.16°, 0.15°, and 0.11°, respectively. The results prove that the proposed method in this paper exhibits high computational efficiency and accuracy, providing an effective approach for digital assembly of mechanical parts.

An efficient infinity norm minimization algorithm for under-determined inverse problems

The problem of solving under-determined systems of linear equations with minimum peak magnitude (ℓ∞ norm) has numerous applications in signal processing. These include Peak-to-Average Power Ratio (PAPR) reduction in MIMO-OFDM systems, vector quantization, approximate nearest neighbor search, optimal control in robotics, and power grid optimization. Several methods have been proposed to address this problem, but they often face limitations in computational speed or representation accuracy. Some methods also impose constraints on the frame matrix, such as restrictions on the type of its entries or its aspect ratio. In this paper, we present the Fast Iterative Peak Shrinkage Algorithm (FIPSA), which iterates over feasible solutions to consistently reduce peak magnitude and provably converge to near-optimal solutions. Our experimental results, conducted across various frame matrix types and aspect ratios, demonstrate that FIPSA consistently achieves near-minimal ℓ∞ norm values. In addition, it operates 1.3 to 7.3 times faster than previous methods, while maintaining an average representation error of 10−15. Notably, these advancements are achieved without imposing any constraints on the frame matrix.

Synchronization Optimization of Pipeline Layout and Pipe Diameter Selection in a Drip Irrigation Network System Based on the Jaya Algorithm

To address the complexity and high computational burden in the design of drip irrigation networks, the Jaya algorithm is utilized to study factors affecting project costs, including equipment and pipeline depreciation and the operation and management costs of the irrigation area. A mathematical model of synchronization optimal design of pipe layout and pipe diameter selection in a drip irrigation network system with constraints on pipe diameter, flow velocity, and pipe pressure is established. Using an irrigation district in Xinjiang, China, as an example, the Jaya algorithm optimization design program was run independently 50 times, and the relative deviation of each optimization result from the optimal solution was calculated. The results show that the annual cost per unit area o is reduced to 635.99 RMB/hm2, a 25.34% reduction compared to the original engineering program, and the investment-saving effect is obvious. The relative deviation is controlled within 3%, which shows that the algorithm has stable convergence performance and can meet the requirements of actual engineering design. The Jaya algorithm eliminates the need for parameter tuning, and it excels in cost savings, algorithm stability, and computational accuracy, making it an effective method for the single-objective optimization design of drip irrigation networks.

Kilometer-Scale Precipitation Forecasting Utilizing Convolutional Neural Networks: A Case Study of Jiangsu’s Coastal Regions

High-resolution precipitation forecasts play a pivotal role in formulating comprehensive disaster prevention and mitigation plans. As spatial resolution enhances, striking a balance between computation, storage, and simulation accuracy becomes imperative to ensure optimal cost-effectiveness. Convolutional neural networks (CNNs), a cornerstone of deep learning, are examined in this study for their downscaling capabilities in precipitation simulation. During a precipitation event on 23 June 2022, in Jiangsu Province, China, distinct rain belts emerged in both southern and northern Jiangsu, precisely captured by a numerical model (the Weather Research and Forecasting, WRF) with a 3 km spatial resolution. Specifically, precipitation was prevalent in northern Jiangsu from 00:00 to 11:00 Beijing Time (BJT), transitioning to southern Jiangsu from 12:00 to 23:00 BJT on the same day. Upon dynamic downscaling, the model reproduced precipitation in these periods with an average error of 12.35 mm at 3 km and 12.48 mm at 1 km spatial resolutions. Employing CNN technology for statistical downscaling to a 1 km spatial resolution, samples from the initial period were utilized for training, while those from the subsequent period served for validation. Following dynamic downscaling, CNNs with four, five, six, and seven layers exhibited average errors of 8.86 mm, 8.93 mm, 9.71 mm, and 9.70 mm, respectively, accompanied by correlation coefficients of 0.550, 0.570, 0.574, and 0.578, respectively. This analysis indicates that for this precipitation event, a shallower CNN depth yields a lower average error and correlation coefficient, whereas a deeper architecture enhances the correlation coefficient. By employing deep network architectures, CNNs are capable of capturing nonlinear patterns and subtle local features from complex meteorological data, thereby providing more accurate predictions during the downscaling process. Leveraging faster computation and reduced storage requirements, machine learning has demonstrated immense potential in high-resolution forecasting research. There is significant scope for advancing technologies that integrate numerical models with machine learning to achieve higher-resolution numerical forecasts.