125 results about "Particle model" patented technology

The invention discloses a numerical simulation method and system for corrosion expansion cracking of reinforced concrete based on near-field dynamics, and relates to the technical field of numerical simulation of civil engineering materials. Discretizing a concrete calculation area into a near-field dynamic material point model to form a discrete model; applying boundary conditions to the discrete model; establishing a mapping relation between the material point damage degree and the erosion coefficient; acquiring chloride ion concentration distribution and oxygen concentration distribution of the discrete model at the current time step, and judging whether the chloride ion concentration of the surface of the steel bar reaches a blunt removal threshold value or not; if the reinforcing steel bar is blunt, obtaining a displacement field of the model at the current time step and the damage degree of each material point, and updating the erosion coefficient of each material point in the model; according to the method, a crack path does not need to be preset, initiation and expansion of complex cracks can be naturally described, the problem of singularity of a crack tip and the problem of convergence do not exist, damage behaviors such as brittle fracture and peeling of concrete can be simulated, and extra constitutive adjustment and model adjustment are not needed.

The invention provides a gas-liquid-solid multi-phase conveying centrifugal pump performance prediction method and an optimization design method. According to the method, the performance of the centrifugal pump is accurately predicted and the design efficiency of the centrifugal pump is improved through numerical simulation and optimization design in combination with a gas-liquid-solid three-phase flow dynamic model. Firstly, blade parameters are determined according to design requirements, a three-dimensional model is established, and grid division is carried out; then, the VOF-Level Set multiphase flow model, the particle model and the gas-liquid-solid coupling model are corrected, factors such as surface tension, particle immersion force and drag force are considered, and lift and efficiency curves are calculated through numerical simulation; and finally, a particle accumulation area is analyzed based on the particle distribution characteristic graph, and the optimal lift efficiency combination is obtained by adjusting parameters such as the inlet angle, the inlet width, the wrap angle and the blade number of the blades and optimizing design. According to the method, the performance of the centrifugal pump in the gas-liquid-solid multiphase flow environment can be improved, and a new design thought is provided.

The invention relates to the field of powder material manufacturing, and discloses a powder particle model parameter calibration method based on a discrete element method.The method comprises the steps that a static accumulation angle experiment is conducted, and the accumulation angle of micron-sized powder is obtained; performing particle size scaling according to a powder particle size distribution curve; discrete element intrinsic parameters are set according to powder characteristics and mold materials; setting to-be-calibrated parameters and ranges; performing PB design according to the to-be-calibrated parameters and the range; analyzing a test result and screening test parameters which have relatively obvious influence on the simplified and scaled powder accumulation angle; establishing a regression model and seeking an optimal combination by taking a relatively significant test parameter as an independent variable and taking an actually measured static accumulation angle as a target value; and finally determining the mineral particle model parameters based on the discrete element method through the established optimization combination verified by a numerical test. According to the method, the parameter calibration efficiency and the parameter calibration accuracy are improved, and an effective technical means is provided for key parameters of a powder particle numerical simulation experiment.

The invention discloses a particle geometric reconstruction and morphological parameter automatic evaluation method based on a three-dimensional image, and the method comprises the steps: firstly obtaining a three-dimensional gray image of a particle sample, carrying out the Gaussian noise reduction, binaryzation and watershed segmentation preprocessing, and achieving the precise segmentation of independent particles based on an established particle segmentation defect repair algorithm; and then, according to the segmented and marked three-dimensional particle image, an initial contour surface is reconstructed by using a marching cube algorithm to generate a particle surface grid, noise is suppressed by using a Taubin smooth optimization algorithm, a topological structure is automatically calibrated, and a digital model conforming to real particle surface features is constructed. And finally, importing each particle model into the created particle morphological parameter automatic evaluation software to realize quantitative analysis of particle morphological characteristics.

The invention provides a construction method and device of a broken particle discrete element numerical model and application of the construction method and device. The method comprises the following steps: constructing N isometric particle contour models based on N real particles; based on the target grading curve and the total number of the particles, constructing a spherical particle set model; replacing each spherical particle model in the spherical particle set model with a breakable model composed of a plurality of polyhedrons one by one; and after all the spherical particle models are replaced by fragmentable models, generating a fragmentable particle material numerical model instruction file, and directly inputting the instruction file into PFC3D in a reading manner to generate a fragmentable particle discrete element numerical model. The breakable particle discrete element numerical model obtained through the construction method has the shape of real particles and is composed of coplanar Voronoi polyhedrons, the size of the model conforms to grading characteristics, and analysis of particle breaking characteristics and the microscomic influence of particle breaking on material macromechanical characteristics is facilitated.

The application discloses a kind of based on three-dimensional particle special effect's airborne chaff bomb simulation method, contain the following steps: chaff individual particle modeling: based on three-dimensional particle special effect technology, simulate the motion trajectory of single chaff, the motion characteristics of chaff cloud and chaff electromagnetic scattering characteristic RCS;Complex environment modeling: including meteorological environment simulation model and electromagnetic environment simulation model;Particle special effect driven simulation calculation: under the action of the chaff individual particle model and complex environment model, first, each chaff individual is particleized;Then real-time calculation is carried out;Finally, radar wave irradiation analysis is carried out;Three-dimensional dynamic visualization output: the distribution and interference effect of chaff bomb under different height and environment are realized Real-time display.The application can highly realistic simulate the explosion, diffusion, disappearance etc. of chaff bomb in launching, including the motion trajectory of chaff bomb, diffusion distribution form and the effect of radar scattering.

The application discloses a kind of experimental measurement methods of non-spherical particle energy recovery coefficient, it is characterized in that, including the following steps: 1) establish the flat plate model of particle-wall collision working condition;2) establish non-spherical particle model and carry out DEM simulation;3) according to the motion parameter of non-spherical particle after collision, combined with free fall motion, the landing point of particle is calculated, and the random landing point distribution of non-spherical particle after collision with wall is quantified using probability density function;4) design several non-spherical particle models to carry out the above simulation and calculation, obtain the energy recovery coefficient and random landing point distribution of several different non-spherical particles;5) build artificial neural network training model, use Python tool to establish the nonlinear fitting relationship between random landing point distribution and energy recovery coefficient by artificial neural network, construct non-spherical particle energy recovery coefficient prediction model;The application has important significance to the research of particle-wall collision field.

The application discloses a complex geometric model particle discrete method and system and electronic equipment, belongs to the field of fluid mechanics, proposes a kind of efficient, accurate method for constructing model geometry surface narrowband area LSDF, and carry out particle compensation, to avoid the problem that particle spacing increases after particle relaxation;According to the above method, by inputting the geometric model and the particle spacing after discretization, the complex geometric model can be discretized into uniform, adaptive particles with specified spacing. The method solves the problem of inaccurate and inefficient LSDF in the prior art, and the volume error of the model after being discretized into particles is large. By inputting the STL geometric model file and the particle spacing, the model can be automatically discretized into a particle model. The particle model can be applied to numerical calculation of the particle method. The accuracy and universality of the method and device in the application are verified by particle discretization of various complex geometries.

The invention relates to a gamma radiation field reconstruction method based on Lagrange particle model parameter assimilation, and the method comprises the steps: determining the spatial positions of a plurality of radioactive aerosol particles based on a Lagrange particle model; constructing state parameters based on the plurality of spatial positions and the source intensity of each aerosol particle; forming a background state vector by using the state parameters, and dynamically updating the background state vector by adopting an ensemble Kalman filtering algorithm to obtain a corrected assimilation state; determining an optimized transportation track based on the spatial position of the radioactive aerosol particles in the corrected assimilation state; constructing a local concentration field corresponding to the radioactive aerosol particles by adopting a Gaussian kernel function; and determining a gamma radiation field corresponding to the radioactive aerosol particles. The three-dimensional radiation field reconstruction is carried out based on the background state vector and the real-time monitoring data by taking the parameters of the radioactive aerosol particles as the core, and the purposes of correcting the background state variable, obtaining the corrected assimilation variable and improving the accuracy of the three-dimensional radiation field reconstruction are achieved.

The application relates to the technical field of particle flow simulation, in particular to a CAE simulation method for C-type particle flow state based on a double Euler model, which defines model parameters under a secondary agglomeration theoretical model according to the agglomeration mechanism of C-type particles from initial agglomeration to secondary agglomeration, the model parameters including particle phase diameter, particle phase density, particle phase volume fraction and particle phase shear viscosity, thereby establishing a local dynamic agglomeration particle model (LDAP model) for C-type particles based on an Euler multiphase flow model, and performing C-type particle flow state simulation; the method can realize the setting of numerical simulation conditions and the simulation of the flow behavior of C-type particles through simple flow field calibration and parameter setting, can intuitively reflect the real-time state of the particle agglomerate diameter at any local position in the flow field, has higher accuracy in describing the motion state of particles and gas under the condition of particle agglomerate diameter difference, and can balance the calculation efficiency without increasing the number of particle phases for varying particle diameters.

A method is provided of simulating additive manufacturing in which powder is used to additively manufacture an object The method comprises using a particle model to model application of powder, for example in form of a deposited powder bed, over a substrate layer during the additive manufacture, initiating a simulation model of the additive manufacturing, wherein the simulation model uses computational fluid dynamics to simulate 3D volumetric fields at respective time instances during the additive manufacturing, and executing the simulation model to obtain at least a 3D phase field and a 3D temperature field at respective time instances during manufacture of a first line structure on the substrate layer to simulate temperature evolution and phase changes during the manufacture of the first line structure.

The application relates to a numerical prediction method for the flow dynamics characteristics of coarse particle solid-liquid two-phase flow in a vertical lifting pipeline, and belongs to the technical field of numerical prediction of flow dynamics characteristics. In order to solve the problem that the influence of particle movement on the surrounding fluid is not considered in the current theoretical calculation and numerical simulation of the solid-liquid two-phase flow in the pipeline, the application firstly models and divides the grid of the fluid domain; a particle model is established based on a particle material simulation software, and the particles are controlled in a discrete manner; then, a coupling interface is established based on a computational fluid dynamics software through a coupling interface control equation; the computational fluid dynamics software calculates and monitors the state of the fluid, which is transported to the particle material simulation software as the external environment of the particles; the interference caused by the movement of the particles is also transmitted to the computational fluid dynamics software through the coupling interface; and the numerical prediction of the flow dynamics characteristics of the coarse particle solid-liquid two-phase flow is realized based on the initialized coupling interface, the pipeline model and the particle model.

The invention relates to the technical field of visual detection of solid waste particles, in particular to an intelligent identification method and system for solid waste particles based on multi-modal fusion, and the method comprises the steps: firstly, carrying out the particle segmentation and particle size statistics of a solid waste infrared image, and calculating a particle adhesion evaluation value; dividing the solid wastes into low, medium and high adhesion grades; then, for different adhesion levels, channel weighted fusion guided by infrared images, or frequency domain fusion guided by RGB images, or instance level alignment fusion based on three-dimensional point cloud space anchor points is adopted to obtain multi-modal fusion features; and finally, the infrared image and the multi-modal fusion feature are mapped to a three-dimensional point cloud model to construct a three-dimensional solid waste particle model, accurate geometric dimensions and material features are obtained at the particle instance level, and the accuracy of solid waste particle type recognition and particle size measurement and the overall recognition efficiency are remarkably improved.

The application discloses a long-time-domain and reaction type combined motion trajectory generation method of a foot-type robot, which comprises the following steps: establishing an equivalent single-particle model of the robot; establishing an equivalent single-particle model cost function based on a center-of-mass acceleration vector and a foot position vector; and establishing a spatial constraint equation of a system state; then, a nonlinear optimization equation is constituted, an optimization solver is used to generate a long-time-domain center-of-mass trajectory and a foot sequence from a starting time to a termination time; a model prediction control and a velocity field guided trajectory are combined to converge the short-time-domain reaction type trajectory to the long-time-domain motion trajectory, and a global motion trajectory is generated. The motion trajectory generation method considers a motion target position, robot kinematics / dynamics and environmental constraints, allows simultaneous foot and center-of-mass trajectory optimization, and still can quickly converge to an expected position when a center-of-mass position deviates from a long-time-domain motion trajectory due to external disturbance, and has the characteristics of strong anti-disturbance ability.

The invention relates to an underground reservoir coal pillar protective layer fracture water inrush large deformation prediction method and system. The method comprises the steps that a particle model containing a coal pillar protective layer, reservoir water flow and a support boundary is generated based on preset particle precision; based on a particle model, a response process of a coal pillar protection layer under a water storage-excavation coupling working condition is simulated through a meshless particle method, and the method comprises the following steps: dynamically updating particle distribution of an excavation area by adopting a particle inactivation algorithm, and simulating a thickness change of the coal pillar protection layer; based on a fluid-structure interaction contact algorithm, calculating the interaction force between the reservoir water flow and the coal pillar protection layer; the number of damaged particles is counted in real time according to the stress state of the coal pillar protective layer particles, and the crack expansion, penetration and fracture processes are identified; and generating a critical protective layer thickness, water inrush outburst time and a fracture influence range based on the damage evolution data, and outputting visual data. Through model construction, dynamic simulation, damage identification and parameter prediction, the technical bottleneck of a traditional grid method in large deformation simulation can be broken through.

The invention provides a glass coarse crushing simulation method based on discrete elements, and belongs to the technical field of glass material processing.The method comprises the steps that firstly, a stacking angle simulation model is built by means of discrete element software on the basis of actual material physical property parameters, and contact parameters between particles are obtained; then, parameters of a BPM (Bonded Particle Model) are defined by comparing a physical experiment with a simulation uniaxial compression result, and finally, the efficient and high-precision DEM simulation method for glass coarse crushing is established according to all the obtained parameters. The constructed model can truly reproduce the whole crushing process under different materials, equipment structures and process parameters, and provides key technical support for deeply understanding the mechanism, optimizing the process, improving the equipment performance and reducing the research and development cost.

The invention discloses a lunar soil coring dynamic process simulation method based on FEM-SPH coupling, and the method comprises the steps: building a three-dimensional geometric model of real lunar soil particles based on the scanning data of the real lunar soil particles; establishing a lunar soil discrete element model based on the three-dimensional geometric model; performing a virtual mechanical test on the lunar soil discrete element model, and calibrating lunar soil macromechanical parameters; constructing a drilling tool finite element model, and defining a contact relation between the drilling tool finite element model and the lunar soil; based on the calibrated lunar soil macromechanical parameters, establishing an SPH particle model of the lunar soil; coupling the drilling tool finite element model and the SPH particle model, and performing lunar soil coring dynamic process simulation based on the contact relationship to obtain bedding information disturbance and drilling tool sampling rate results. According to the method, accurate simulation of the dynamic process of lunar soil coring is achieved, and a sampling rate and bedding information disturbance quantitative basis is provided for optimization of drilling tool parameters and drilling working conditions.

The invention relates to a method for rapidly solving gasification and combustion of single-particle biomass coke, and belongs to the technical field of biomass thermochemical conversion simulation. In order to solve the problem that the reaction rate, the heating rate and the mass transfer process are overestimated due to the fact that a traditional zero-dimensional model neglects mass transfer and heat transfer resistance in particles, the difference between the zero-dimensional model and a one-dimensional model in the aspect of transfer behaviors in the particles is compared and analyzed through a system; a mass transfer rate correction factor (HM), a heat transfer rate correction factor (HT) and a reaction rate correction factor (HR) are introduced to construct a corrected zero-dimensional particle model. According to the model, on the basis that the calculation efficiency advantage of a zero-dimensional model is reserved, the prediction precision of the particle internal transfer effect and the influence of the particle internal transfer effect on the reaction process is remarkably improved. Besides, the method can be efficiently coupled with a computational fluid dynamics (CFD) method, an effective tool is provided for realizing high-precision and high-efficiency numerical simulation of the biomass gasification and combustion reactor, and a theoretical basis is laid for engineering amplification and operation optimization of the biomass gasification and combustion reactor.

The present application belongs to the field of geothermal energy development and geological engineering technology, and discloses a geomechanics-constrained geothermal reservoir fracture network dynamic modeling method. The method comprises the following steps: first, obtaining multi-source monitoring data of the geothermal reservoir, including rock mass surface deformation data, micro-damage spatiotemporal distribution data and acoustic emission rupture signal data; constructing a micro-macro correlation model based on the bonded particle model and the tensor theory to obtain the eigenvalues of rock mass damage evolution; positioning the three-dimensional rupture source of the acoustic emission rupture signal; constructing a cubic cluster model based on the positioning results to generate the initial topological structure of the fracture network; dynamically adjusting the geometric shape and connectivity of the fracture network through the random walk algorithm and the fracture intersection correction mechanism according to the evolution eigenvalues and the geomechanics constraint conditions; and finally predicting the fracture expansion trend and damage evolution law under different geomechanics conditions. The present application significantly improves the accuracy and reliability of fracture network evolution prediction.

The application provides a modeling method of a real particle shape breakable particle model based on controllable grid division quantity and regularity degree, belongs to the technical field of the breakable particle model, and comprises the following steps: firstly, photographing, extracting and scanning a real particle shape to obtain particle contour information; then, reading the particle contour information through an editor, generating random points and circumscribed rectangles in the contour by using a control grid regularity degree method, exporting circumscribed rectangle information of the random points, importing random point coordinate information into neper software to perform Voronoi division on the circumscribed rectangles, exporting the divided unit set information, again importing the unit set information into a python editor to cut and retain the rectangles by using the particle contour, and finally assembling the processed Voronoi units into a breakable particle model. Based on real particle data obtained by photographing or scanning, the modeling method of the cohesive cell method is used to realize the modeling of the real particle shape, so that the model can more accurately simulate the shape, surface roughness, volume and other parameters of the real particle.

The invention relates to the technical field of battery power prediction, and discloses an electrochemical model SOP prediction method based on a physical information neural network, and the method comprises the following steps: constructing an enhanced single particle model, taking an ordinary differential equation and a heat conduction equation of the enhanced single particle model as constraints, and embedding the constraints into a physical information neural network model; according to the prediction method, data generated by a P2D electrochemical model is introduced through an offline training stage, a control equation is embedded into a neural network loss function in a residual form, the control equation comprises lithium ion diffusion, charge conservation and Butler-Volmer reaction kinetics, and the physical constraint enables the PINN not only to be accurate in fitting result, but also to be high in precision and high in reliability. And moreover, the evolution law of key state variables in SOP estimation can be better captured, high precision is achieved even under the complex dynamic condition, and the model is remarkably superior to a pure data driving model.

The invention discloses a method and a system for calculating the mixing proportion of multi-end-member water mass mixed particles. The method comprises the following steps: calculating the contribution proportion of each water mass end member of a mixing area by adopting an isotope mixing model; generating initialization information of a multi-end-member water mass mixed particle model by using the delta < 18 > O and salinity of each end member and derivative information; based on the contribution proportion and the initialization information, constructing a multi-end-member water mass mixed particle model; simulating a simulated delta < 18 > O-salinity linear relation of a mixing area under all iterative mixing proportion combinations based on a multi-end-member water mass mixed particle model; and calculating the mixing ratio of each end member by comparing the simulated delta 18O-salinity linear relationship with the actual delta 18O-salinity linear relationship. According to the method, the multi-end-member water mass mixed particle model is constructed, and iterative calculation and fluctuation simulation are combined, so that accurate quantification of a water mass mixing ratio and quantitative analysis of mixing sufficiency are realized.

This invention provides a method, apparatus, electronic device, and medium for determining contact force in particle motion simulation, belonging to the field of particle simulation. The method includes: acquiring a particle model to be simulated, the particle model including multiple nodes; performing surface reconstruction processing on the surface of the particle model to obtain a reconstructed surface; setting multiple background points on the reconstructed surface; for a target particle model, determining the contact force generated by other particle models on the target particle model based on the background points of the reconstructed surface of the target particle model and the nodes of other particle models; and performing particle motion simulation on the target particle model based on the contact force. This invention improves computational efficiency while ensuring the accuracy of particle simulation by performing surface reconstruction on the mesh surface of the particle model to obtain a smooth reconstructed surface, and then setting multiple background points on the reconstructed surface. The contact force is solved by the relationship between the background points and nodes.

The application provides a numerical method for simulating fiber-containing proppant flowback, and the method comprises the following steps: establishing a simplified fracture network grid model for flowback, setting a fracture fracturing fluid inlet and a fracturing fluid outlet; setting a proppant particle model, using a Bonding V2 model to establish a flexible fiber; according to a set mass ratio, stacking the fiber and the proppant in the fracture; in a finite element software, setting a flow equation, fracturing fluid material properties and fracture inlet and outlet flow rate parameters; reading a coupling file, starting to simulate flowback, and post-processing the calculation results. Through the method, the sand stabilizing and sand fixing capacities of different fiber lengths, different fiber concentrations and different fiber mixing ratios can be studied, and the method has the advantages of simple operation and high calculation efficiency.

The invention belongs to the technical field of methods for simulating and evaluating the wear resistance of a screw, and particularly relates to a method for simulating and evaluating the wear resistance of the screw based on discrete elements, which comprises the following steps of: constructing a three-dimensional assembly model of the screw and a machine barrel, importing the three-dimensional assembly model into discrete element simulation software EDEM, and finishing parameter setting in the EDEM, a particle model is created through a particle generator, the density and Poisson's ratio of particles, the static / dynamic friction coefficient with the surface of a screw and particle size distribution are defined, the density, shear modulus, Poisson's ratio and friction coefficient of the screw and a machine barrel are input, and the contact model is used for describing mechanical behaviors between the particles and between the particles and the screw. The abrasion model is used for representing the screw abrasion process, determining the size and the feeding efficiency of a virtual feeding factory, adjusting the rotating speed of the screw according to equipment parameters, configuring simulation calculation parameters and then starting simulation calculation.

The application discloses a kind of liquid metal particle filter clogging discrete element and macroscopic particle model coupling simulation method, establishes the geometry model of flow calculation domain inside filter and divides grid;Calculate fluid velocity field, adopt discrete element method to track the motion of particle in fluid and the interaction between it and fluid, calculate the collision, adhesion and deposition behavior between particle and particle, particle and wall surface;Adopt macroscopic particle model to characterize the feedback effect of particle volume on local flow field, update velocity field;The method can realize the simultaneous simulation of important variables such as particle behavior and flow field change during filtration process, realize the dynamic characterization of particle migration, capture, clogging and filter cake formation process, and can provide calculation basis for filter structure design and operation optimization.

The embodiment of the application discloses a kind of high-voltage circuit breaker spring fatigue degree evaluation methods, the method comprises: the particle model of high-voltage circuit breaker is constructed, the particle model includes several to be measured particles;The vibration signal of the several to be measured particles is collected;According to ensemble empirical mode decomposition, the vibration signal is decomposed, obtains several to be measured vibration signals;According to the vibration energy of the several to be measured vibration signals and its corresponding weight coefficient, the spring energy storage function corresponding to the several to be measured particles is determined;According to the spring energy storage function, the energy storage detection value of high-voltage circuit breaker spring is determined, and the energy storage detection value is used to characterize the fatigue degree of spring.The application is realized to the accurate evaluation of high-voltage circuit breaker spring fatigue degree by constructing particle model, collecting vibration signal and decomposing, determining spring energy storage function, not only improves the accuracy and reliability of evaluation, also provides effective technical support for the maintenance and management of circuit breaker.

This invention relates to the field of combustion dynamics technology, and provides a method for modeling the combustion dynamics of nano-aluminum particles based on molecular reaction dynamics. The method includes: S1 constructing a nano-aluminum particle model; S2 establishing an oxidation combustion reaction system; S3 analyzing molecular trajectories; S4 extracting reaction paths; S5 determining gas-phase reaction kinetic parameters; S6 determining surface reaction kinetic parameters; S7 constructing a heterogeneous combustion kinetic mechanism; and S8 verifying and outputting simulation results. This invention directly extracts reaction paths through molecular dynamic trajectories in the reaction force field, reducing the subjectivity of relying entirely on artificially preset elementary reactions. It supplements key gas-phase and surface reaction parameters through quantum chemical calculations and density functional theory calculations, improving the physicochemical basis of the mechanism parameters. By coupling gas-phase and surface reactions into a heterogeneous combustion kinetic mechanism, it more completely describes the structural evolution process of aluminum nanoparticles from melting, rapid oxidation, partial oxidation to equilibrium.

This invention belongs to the field of visual perception technology and provides a method and system for detecting contrast sensitivity functions. It generates a target set by combining a hybrid contrast sensitivity function model of a parameter domain particle model and a stimulus domain grid model. Subjects are then tested using this target set, and the hybrid model can be updated based on the subjects' responses. The updated model then generates new target sets for further testing. This approach employs a hybrid contrast sensitivity function model and a resampling strategy. The hybrid contrast sensitivity function model simultaneously provides parameter estimation functionality in the parameter domain and records the information accumulation of each stimulus point target in the stimulus-target domain. This helps generate more effective test targets for subjects and improves the method's adaptability to individual differences. The resampling strategy improves test resolution without increasing computational load, ensuring high accuracy under various conditions.

The application provides a training set construction method for noise robustness enhancement and related devices, the method comprising: obtaining a target multi-view image of a target object and a target object mask image in a robot noise operation scene; determining a reference particle set according to the target multi-view image and the target object mask image; performing parameter configuration and noise injection on the reference particle set to obtain a target particle model a; screening the target particle model a to obtain a target particle model b, and inputting the target particle model b into a differentiable particle engine for noise scene simulation to obtain b groups of target noise data; screening the b groups of target noise data to obtain c groups of target noise data; and determining a target training set according to the c groups of target noise data and original noise-free data. Through controllable noise injection, physical simulation and effectiveness screening, a training set containing high-quality noise samples is constructed, and the noise robustness of the model is improved.