508 results about "Numerical computing" patented technology

The invention discloses a coupling iterative algorithm library and a reactor multi-physical field coupling calculation program construction method. The coupling iterative algorithm library comprises an open source library layer used for integrating a plurality of numerical calculation libraries and providing an input / output interface for a reactor multi-physics field coupling framework and each numerical calculation library; the core algorithm layer is used for integrating a plurality of coupling iteration algorithms and a plurality of acceleration algorithms related to each coupling iteration algorithm; the interface layer is used for providing a data interaction interface between a nuclear reactor multi-physics field coupling calculation program and the core algorithm layer; and the iterative algorithm template layer is used for forming an iterative solution process control template function based on the open source library layer, the core algorithm layer and the interface layer, and the iterative solution process control template function is used for constructing a nuclear reactor multi-physics field coupling calculation program based on the process control parameters set in the input file. According to the coupling iterative algorithm library, the development efficiency and the development quality of a nuclear reactor multi-physics field coupling calculation program can be improved.

The invention relates to the field of reactor data processing, and particularly discloses a numerical calculation method for solving a neutron transport equation. The method comprises the following steps: 1, describing the geometry of a reactor core of a nuclear reactor by utilizing a construction entity geometric modeling method, and subdividing a curved surface grid along a curved surface structure; 2, directly expanding a neutron flux distribution function in each curved surface grid by using a multivariate polynomial; 3, deriving a neutron transport equation and an integral residual equation of a boundary condition under each curved surface grid; 4, solving geometric integral parameters of each curved surface grid; step 5, establishing an adaptive linear system; and step 6, iteratively calculating and solving a maximum characteristic value and an unknown expansion coefficient.

The invention discloses a comprehensive intelligent monitoring decision-making system and method for an overlapping shield tunnel system. The system comprises an information acquisition system, an iterative optimization state real-time evaluation system and an intelligent maintenance decision-making system. Firstly, a preset model is trained through data such as numerical calculation; operating measured data to train a feedback model, and weighting output results of the two models to obtain a tunnel deformation predicted value; then establishing an experience overlapping tunnel safety evaluation system and an iterative optimization evaluation system, performing deep association on the tunnel state, damage and maintenance decision through association degree intelligent real-time analysis, realizing tunnel state real-time monitoring, tunnel damage real-time capturing and tunnel maintenance decision real-time updating, performing intelligent matching on maintenance measures, and performing real-time monitoring on the tunnel state, the tunnel damage and the maintenance decision. And the prediction model and the safety evaluation system are iteratively optimized according to operation effect feedback. According to the invention, real-time perception, intelligent diagnosis and maintenance of the whole life cycle overall service performance of the overlapping shield system are realized, and the long-term service quality of the overlapping shield tunnel structure is improved.

The invention provides a computing power metering method and device of an intelligent computing center cloud platform based on a multi-dimensional performance index, and relates to the technical field of intelligent computing centers, intelligent computing centers, computing power infrastructures and intelligent computing clouds, and the method comprises the steps: S1, obtaining a first task type of a first computing power operation task; s2, determining a plurality of resource indexes corresponding to the first task type and a weight coefficient corresponding to each resource index; s3, collecting a parameter value corresponding to each resource index in each of the N periods; s4, calculating a computing power value corresponding to the first computing power operation task based on the weight coefficient corresponding to each resource index and the parameter value corresponding to each resource index in N periods; and S5, generating a first verification data packet based on the first task identifier of the first computing power operation task, the weight coefficient corresponding to each resource index and the parameter value corresponding to each resource index in N periods. The computing power credibility of the intelligent computing center cloud platform can be greatly improved.

The invention relates to the technical field of underground engineering safety monitoring and disaster simulation, in particular to a digital twinborn deduction platform for underground engineering disaster chain evolution simulation, which comprises a multi-disaster coupling numerical simulation module used for acquiring multi-source data including design drawing data, geological data and engineering monitoring data, and based on the multi-source data, establishing an initial stress field model by using a numerical calculation method combining a finite element and a finite difference, and carrying out leakage-settlement-structure failure multi-disaster coupling numerical simulation. According to the method, the partial differential equation for describing the evolution law of the underground engineering is used as a constraint term to be embedded into the deduction model, and online identification and dynamic extrapolation of parameters are executed by fusing real-time monitoring data, so that the accuracy of disaster evolution simulation under complex working conditions is remarkably improved; and the problem that the traditional numerical calculation method is difficult to meet the real-time requirement of digital twinning is solved.

The invention discloses a data center temperature field and flow field rapid order reduction simulation method and system, and the method comprises the following steps: S1, carrying out the CFD numerical calculation of a data center, and obtaining a CFD data set; s2, performing dimensionality reduction on the CFD data set by adopting a weighted intrinsic orthogonal decomposition (POD) method to obtain a low-dimensional mode; s3, optimizing the low-dimensional mode by adopting a machine learning model to obtain a reduced-order model; s4, performing error evaluation and physical constraint processing on the reduced-order model, and performing verification; and S5, generating a temperature cloud chart and a streamline chart by adopting a reduced-order model, and performing hotspot optimization based on a visualization result. Temperature field and flow field simulation is carried out under the guidance of the thermodynamic theory, and the actual requirement for efficient heat management of the data center is met.

The invention discloses a flow field prediction-oriented full-process automatic joint reduced-order modeling method, which comprises the following steps of: specifying a target physical field parameter space, and randomly generating a sample space according to a Latin hypercube sampling method; constructing a full-process automatic simulation tool chain, driving target physical field numerical calculation and generating a training data set; carrying out singular value decomposition-based intrinsic orthogonal decomposition on the output physical field data, and only retaining first r main feature components to construct a reduced-order data set; constructing a multi-input multi-output full-connection feedforward neural network, and modeling and training a nonlinear mapping relation between input parameters and reduced-order features; new working condition parameters are input, reduced-order features are predicted through the trained neural network, distribution of a target physical field is reconstructed according to a singular value decomposition reduction matrix, and more flexible and reliable technical support is provided for reducing the training cost of a reduced-order model and improving simulation efficiency.

The invention discloses a rock fracture prediction method and system based on a resistivity data assimilation algorithm, and the method comprises the steps: building a numerical model for simulating the asymptotic fracture process of a rock mass under the load effect according to the boundary conditions and initial conditions of the rock mass; simulating the stress damage process of the rock mass by using a numerical calculation method to obtain prediction data of stress field and fracture damage field distribution of the rock mass; monitoring the damage process of the rock mass sample in real time by using a high-density electrical method to obtain observation data; rock mass fracture damage distribution is obtained based on observation data; an assimilation algorithm is adopted, information of observation data and information of prediction data are combined, and initial conditions or parameters of the numerical model are corrected; and then corrected rock fracture prediction data is obtained. Through the data assimilation algorithm, different data and model simulation results are fused in the power frame of rock fracture, the model track is automatically adjusted continuously depending on observation, the analysis result with higher precision and more consistency is obtained, and the prediction precision and predictability of the model are improved.

The invention relates to the technical field of tunnel deformation prediction, in particular to a BPNN parameter inversion and LSTM tunnel deformation prediction method. On the basis of the orthogonal test design principle, multiple parameter combinations are designed for the M-C constitutive model and the HSS constitutive model, numerical calculation is carried out, and vertical displacement data of the tunnel measuring points are obtained through numerical calculation; constructing a BPNN model, and training the BPNN model by taking the vertical displacement data as input and the corresponding constitutive model parameters as output; inputting field actual measurement tunnel settlement data into the trained BPNN model, and performing inversion to obtain optimal constitutive parameters of M-C and HSS constitutive models; performing forward numerical simulation on the whole process of pipe jacking construction by adopting the optimal constitutive parameters obtained by inversion to obtain a tunnel vertical displacement evolution rule; and inputting a vertical displacement monitoring value of the measuring point of the downlink tunnel based on the long-short term memory neural network to obtain a prediction result of future settlement of the tunnel. The accuracy of the predicted value is improved.

The invention belongs to the technical field of computational electromagnetics, and discloses a numerical calculation method for simulating electromagnetic wave propagation characteristics in time-varying plasma, which is a three-dimensional Newton-HIE-FDTD method for efficiently simulating propagation of electromagnetic waves in time-varying plasma based on a hybrid explicit-implicit finite difference time domain method and Newton's second law of motion. The method inherits the advantages of the Newton-ADE-FDTD method, the time-varying electron density is accurately represented through the time derivative, and the abundant dynamic characteristics of the time-varying plasma can be accurately captured. On the basis, the method also fuses the advantages of an HIE-FDTD method, so that the time step length of the algorithm is only restricted by the sizes of grids in two directions, and compared with other simulation methods based on time-varying plasma, the method provided by the invention is more excellent in calculation efficiency and is more suitable for large-scale popularization and application. The method is especially suitable for analysis and simulation of electromagnetic wave propagation characteristics in time-varying plasma.

The invention relates to the field of multiphase numerical simulation, and discloses a large-scale particle flow numerical calculation method based on multi-ball filling, and the method comprises the steps: obtaining a fluid domain grid size, a fluid density and a large particle radius; calculating the radius of the small filling ball according to the size of the fluid domain grid; based on golden angle spiral distribution and minimum distance coverage judgment, determining the number and the positions of small filling balls, and generating a large particle model of multi-ball filling; obtaining the background flow field speed of the large particles and the filling small balls in the current time step, and calculating the background flow field speed and drag force of the large particles; updating through iterative calculation to obtain the speed and the position of the large particles and the small filling balls in the next time step; and adding drag force as a source item into a momentum equation to solve fluid motion, and obtaining a fluid computational domain grid speed in the next time step. Accurate numerical simulation of large-scale particles is realized.

The invention belongs to the technical field of nuclear reactor physical numerical calculation, and particularly relates to a method for improving precision and convergence of solving a diffusion equation based on a PINN, and the method comprises the following steps: S1, building a neural network algorithm model combining a physical information neural network PINN and source iteration; s2, aiming at a multi-region smooth transition neutron diffusion problem with small cross section difference between materials, optimizing the neural network algorithm model by adopting an optimization strategy of a training level; and S3, aiming at the multi-region neutron diffusion problem of large neutron flux gradient at the boundary due to large cross section difference between materials, improving the neural network algorithm model by adopting a domain-divided neural network architecture. According to the method, the problem that the calculation cost is exponentially increased in a high-dimensional or complex geometric scene due to the fact that a traditional finite difference method and a finite element method depend on grid division is avoided.

The invention provides a method and a device for dynamically metering computing power based on computing power operation tasks of an intelligent computing center cloud platform, and relates to the technical field of intelligent computing centers, intelligent computing centers, computing power infrastructures and intelligent computing cloud. The method for dynamically metering the computing power of the intelligent computing center cloud platform based on the computing power operation task comprises the following steps: S1, acquiring a parameter value of first operation data of a first GPU in a first time period, wherein the first GPU is a GPU for operating a first computing power operation task; s2, determining a first task type of the first computing power operation task based on a parameter value of the first operation data; s3, determining a weighting coefficient corresponding to the first task type; and S4, based on the weighting coefficient and the parameter value of the first operation data, calculating to obtain the calculation power value of the first calculation power operation task in the first time period. According to the method, the computing power metering accuracy of the computing power operation task in the intelligent computing center cloud platform can be greatly improved.

The invention discloses a semiconductor device critical breakdown characteristic simulation method based on a hybrid finite volume method. The method comprises the following steps: firstly, carrying out geometric modeling on a device, and carrying out mesh generation; then, the quality of the grids is judged, an FBM method is adopted for the grids with the good quality, and a CVFEM method is adopted for the grids with the poor quality to discretize a drift-diffusion equation set; and then, solving the discrete drift-diffusion equation set system by adopting a full-coupling Newton iteration method, and realizing the optimization of the boost numerical simulation process by adopting an automatic boost strategy and an initial value extrapolation method. And finally, obtaining physical field information of a power semiconductor device (LDMOS) under different external bias voltages, calculating terminal current through a post-processing method, and obtaining an I-V characteristic curve of the device after the avalanche breakdown effect is considered. The method has good numerical calculation precision, numerical stability and grid adaptability, and the complete benchmarking with the solution precision of an international advanced semiconductor simulation tool Sentaur TCAD is realized.

The invention belongs to the technical field of heat exchangers, and discloses a step-by-step iteration multi-objective optimization design method and system for a heat exchanger, numerical calculation verification is carried out on performance indexes of decision points, the maximum error of a result obtained through first optimization design is 25.3%, and the maximum error of a result obtained through second optimization design is only 0.45%. Calculation shows that the step-by-step iteration optimization strategy effectively screens out key optimization variables and gradually reduces the optimization range, and multi-parameter, wide-range, low-cost, high-precision and multi-objective optimization design is achieved. From the three performance indexes of the unit volume heat exchange amount, the average Nusselt number and the average friction coefficient, the performance of the large flow and the small fins is better; the heat exchange amount of unit volume can be improved by reducing the transverse spacing of the fins, and the friction coefficient can be reduced by increasing the transverse spacing of the fins. The method and the conclusion provide a new design calculation method and theoretical support for design and optimization of the heat exchanger.

The invention discloses a fracturing fracture extension simulation method based on a self-adaptive multi-scale phase field model, and relates to the field of oil and gas reservoir renovation, and the method is characterized in that reservoir physical property and rock mechanics parameters are obtained, and a reservoir geologic model is established according to the parameters; dividing the reservoir geometric model into a plurality of finite element grids, and assigning grid attributes of the finite element grids by using the parameters to obtain preset hydraulic fractures in the finite element grids and initial variable values on nodes of the finite element grids; according to a preset position relation between the hydraulic fracture and the finite element mesh, constructing a multi-scale adaptive mesh for dynamic expansion of the hydraulic fracture; establishing a coupling mathematical model of a stress field, a seepage field and a phase field of hydraulic fracturing, and constructing a numerical calculation expression of the coupling mathematical model according to the multi-scale adaptive grid; and based on the numerical calculation expression, solving the coupling mathematical model by taking parameters of reservoir physical properties and rock mechanics and boundary conditions of the multi-scale adaptive grid as constraints to obtain hydraulic fracturing fracture extension simulation results under different reservoir and construction conditions.

The present invention relates to the technical field of fluid mechanics. Disclosed is a variable-density low-speed flow mesoscopic numerical calculation method, comprising: determining a computational domain, and dividing or reading grid information, fluid parameters and control parameters in the computational domain; on the basis of the read information, allocating arrays, initializing corresponding parameters of the computational domain, and entering a time step loop upon completion; setting boundary conditions, and assigning all boundary macroscopic quantities in a flow field; under a finite volume framework, deriving a numerical flux of the flow field by means of a BGK equation to solve a Navier-Stokes equation of the flow field; calculating pressure and velocity fluxes at cell interfaces of the flow field, and updating boundary macroscopic quantity information on the basis of flux information; determining whether there is an external force, and if so, applying the external force by means of splitting, and updating the boundary macroscopic quantities; determining whether a time step meets a loop condition; if so, re-entering setting boundary conditions, and assigning all boundary macroscopic quantities in the flow field; and if not, outputting flow field data, and ending simulation. The present invention has the advantages of high algorithm accuracy and simple programming.

The invention relates to a numerical calculation method and device for simulating slippage of a duct piece and a built-in steel bar. The numerical calculation method for simulating slippage of the duct piece and the built-in steel bar comprises the following steps: performing grid division on a three-dimensional shield duct piece according to reinforcement information, endowing the grid three-dimensional shield duct piece with basic physical parameters of concrete, and constructing cable units arranged along actual reinforcement positions, coupling with the concrete model in combination with the calibrated bonding parameters to obtain a concrete model with slippage characteristics; a load path conforming to an engineering scene is applied to the concrete model with the slippage characteristic, nonlinear increment iteration solving is adopted, and simulation time history data is obtained; and finally, performing index calculation and evaluation on the simulation time history data. According to the numerical calculation method for simulating slippage of the duct piece and the built-in steel bar, the problem of overestimation of rigidity and bearing capacity caused by assumption of complete bonding is effectively avoided.

The invention discloses a floating fan aerodynamic load real-time mixing test method and system based on AI enhancement, and the method comprises the steps: calculating the six-degree-of-freedom motion information of a floating foundation through a numerical calculation model based on a real-time floating foundation kinetic equation; the method comprises the following steps: carrying a reduced-scale fan model on a six-degree-of-freedom motion platform, placing the six-degree-of-freedom motion platform on a wind tunnel atmospheric boundary layer, controlling the six-degree-of-freedom motion platform based on floating foundation six-degree-of-freedom motion information, driving the reduced-scale fan model to move, generating an aerodynamic load in a simulated wind environment, measuring an original load signal, and correcting the original load signal. Synchronously recording the actual pose of the six-degree-of-freedom platform; and AI correction is carried out on the corrected load signal and the actual pose of the six-degree-of-freedom platform to obtain a net aerodynamic load, the net aerodynamic load is fed back to the numerical calculation model, and closed-loop updating is carried out to realize coupling circulation. According to the method, the problem of interaction between numerical values and physical boundaries is solved, and the test precision of the large floating fan under the complex environment load can be remarkably improved.

The invention provides a novel finite difference time domain (FDTD) method based on a Newton second law of motion and a bilinear transformation (BT) technology. The method is used for simulating propagation of electromagnetic waves in time-varying magnetized plasma. According to the method, the time-varying electron density is accurately represented through the time derivative, so that abundant dynamic characteristics of the time-varying magnetized plasma can be captured more accurately. A numerical result shows that compared with a Newton-ADE-FDTD method, a traditional ADE-FDTD method and a BT-FDTD method, the Newton-BT-FDTD method provided by the invention has higher precision and calculation efficiency on the time-varying magnetized plasma, especially under the condition of the time-varying magnetized plasma with dramatic electron density change. In addition, the propagation characteristics of the electromagnetic waves in the time-varying magnetized plasma are analyzed.

The invention discloses a flood storage and detention area flood diversion process simulation method and device based on a hydrodynamic model, and the method comprises the steps: obtaining the landform, river section, dike engineering parameters, and real-time hydrological and meteorological data of a flood storage and detention area, constructing a basic parameter library, and storing the basic parameter library in a data input module; constructing a hydrodynamic model calculation module by taking a hydrodynamic equation as a theoretical basis on the basis of the basic database, and combining a finite element volume method; receiving a numerical calculation result, analyzing real-time parameters through a dynamic simulation module so as to match actual hydrological parameters, and performing comparison under multiple scenes; and transmitting the hydrological parameters and the comparison data to a result output module, and outputting key parameters of the flood diversion process and a risk assessment result. The method has the advantages that high-frequency coupling of a hydrodynamic model, multi-source dynamic data, a cross-domain model and an engineering control system is achieved, and the problems of data lag, model isolation and control disjunction in traditional simulation are solved.

The invention relates to the technical field of random geological modeling and numerical simulation integration, in particular to an automatic shield numerical simulation method and device considering geological uncertainty, and the method comprises the steps: extracting geological interface position information based on a pre-constructed voxelization three-dimensional geological model; obtaining structured construction information, and constructing a building informatization BIM model by using the geological interface position information and the construction information; key information of the BIM model is extracted, and a computational grid model and a computational instruction file are respectively generated according to the key information, so that the purpose of calculation integration of the BIM and the numerical simulation software is achieved. Therefore, the problem that according to a design method in the related technology, due to the fact that random geological modeling and numerical calculation analysis are separated from each other, geological uncertainty analysis in shield engineering is difficult to efficiently complete is solved.

The invention discloses a decoy projectile simulation flow field modeling method based on PINN-U-Net and local-global feature fusion, and the method comprises the steps: constructing a simulation data set of a decoy projectile flow field through numerical simulation, and building a physical loss function meeting a decoy projectile temperature field control equation and boundary conditions; designing a Transform-based global feature extraction module, splicing continuous multi-frame time sequence physical quantity sequences in the data set, adding space coordinate information to form an input vector, inputting the input vector into the global feature extraction module, and extracting and outputting a feature vector corresponding to each physical point; replacing the main body network of the PINN with the U-Net network structure to construct a PINN-U-Net network, and splicing the frame data with the corresponding feature vector of each point of the corresponding frame to form an input tensor of the PINN-U-Net network; under the constraint of a physical loss function, outputting a physical quantity in a decoy projectile flow field; according to the method, the modeling speed and calculation precision of a decoy missile combustion scene are remarkably improved, various limitations of traditional numerical calculation are avoided, and a high-speed and accurate scene modeling function can be provided.

The embodiment of the invention provides a spectrum red shift measurement method, device and equipment and a medium, which can be applied to the fields of astronomical information technology and high-performance computing technology, and the method comprises the following steps: receiving N observation spectrum data to be measured; determining a processing mode of the observed spectrum data; in response to the processing mode being a first processing mode, processing the observation spectral data through a first calculation path, the first calculation path performing red shift measurement calculation based on a first numerical calculation library by matching the observation spectral data with a set of template spectral data to obtain a calculation result, the calculation result of the first numerical calculation library is consistent with the calculation result of a reference algorithm within a preset precision range; in response to the processing mode being a second processing mode, the observed spectral data is processed through a second computational path that reconstructs the redshift measurement calculations into a tensor parallel model based on a second library of numerical calculations and performs on parallel computing hardware, the second library of numerical calculations supporting tensor calculations.

The invention discloses a batch modeling method for slope excavation support discontinuous numerical calculation. The batch modeling method comprises the steps that a lithologic model reflecting material partitions is established according to terrain and stratum lithologic characters contained in a slope design result; according to the fracture and joint distribution characteristics, a fracture model and a joint model are established in combination with the calculation purpose, and the inheritance relation among the fracture model, the joint model and the fracture model is established; selecting one of the three models as a parent model, setting an excavation sequence of each bench according to a design result, and constructing a step-by-step excavation model; according to the relative relation between the supporting opportunity and the excavation sequence, corresponding supporting models are created on the basis of the excavation models of all the stair flights in sequence; and according to the inheritance relationship, automatically generating general command stream files containing excavation information, support information and parent model information in batches, and converting the general command stream files into command stream files meeting calculation requirements, so as to establish slope excavation support discontinuous numerical calculation models in batches.

The embodiment of the invention discloses an abnormal bidding behavior recognition method and device, electronic equipment and a storage medium, and relates to the technical field of computers, and the method comprises the steps: determining a bidding enterprise of a target bidding project; calculating historical bidding anomalies of the bidding enterprise, wherein the historical bidding anomalies comprise a huddling risk parameter value, an accompanying bidding risk parameter value and a bidding behavior mutation parameter value; calculating bidding document tamper-resistant similarity of the bidding enterprises; and determining an abnormal bidding behavior identification result of the bidding enterprise according to the historical bidding abnormality and the bidding document tamper-resistant similarity. According to the technical scheme provided by the embodiment of the invention, the problems of low manual auditing efficiency, poor identification accuracy and insufficient multi-dimensional analysis in the prior art can be solved.

The present invention discloses a material mechanical parameter identification method based on interval optimization and indentation testing, which specifically comprises the following steps: Step 1: performing multiple Berkovich indentation tests on the metal material to be tested, obtaining load and displacement curves of multiple loading processes of the nanoindentation test of the tested material, and obtaining the upper and lower limits of the loading curvature C; Step 2: establishing a design space for the material plasticity parameters, and establishing a simulated indentation displacement snapshot matrix X. sim Step 3: Establish an artificial neural network including an input layer, two hidden layers, and an output layer connected in sequence, and optimize the hyperparameters of the established artificial neural network; Step 4: Establish an uncertain optimization problem; Step 5: Based on the interval optimization and interval order relationship of the minimization problem, the uncertain optimization is converted into a deterministic optimization, and a two-layer genetic algorithm nested solution method is used to solve the uncertain optimization problem. The method of the present invention reduces the cost of parameter identification calculation and improves the efficiency of numerical calculation.

The invention relates to the technical field of numerical calculation and data analysis, in particular to a data interpolation and fitting analysis system which comprises a data processing module, a depth generation interpolation module, an uncertainty quantification module, an interpretability optimization module, a multi-source heterogeneous data processing module and a hybrid calculation acceleration module. In the prior art, a traditional interpolation method is easy to generate over-fitting or under-fitting in a complex data distribution and high noise scene, and a single deep learning model is insufficient in generalization ability in a small sample or data sparse region; through the dynamic fusion architecture of the depth generation interpolation module, the advantages of a traditional numerical method and deep learning are combined, the weight is automatically adjusted based on data characteristics, the prediction error under noise data is remarkably reduced, the adaptability to complex distribution is improved, and the stability of a data sparse region is enhanced; the problems of insufficient precision and weak generalization ability of a single model of a traditional method are effectively solved.

The invention relates to the technical field of computational simulation, in particular to a modeling method for constructing any chemical short-range ordered multi-principal element alloy structure, which comprises the following steps: setting target values and convergence conditions of chemical short-range ordered parameters; calculating parameter values of chemical short-range order parameters of the initial configuration; calculating an initial difference value; judging whether the initial difference value meets a convergence condition or not; if yes, outputting an atomic configuration; if not, carrying out iterative operation; the iterative operation process comprises the following steps of: randomly exchanging two atomic positions to obtain a new configuration; calculating a difference value between a parameter value number of the chemical short-range order parameter of the new configuration and a target value to obtain an exchanged difference value; judging whether the difference value after exchange is smaller than the difference value before exchange or not; if yes, current exchange is accepted; and after the convergence condition is met, outputting the atomic configuration. According to the method, the function of directly generating a specific structure from structural features is achieved, and atom configuration of any chemical short-range ordered feature can be customized in a wide component space.

The application discloses a neural network system, a method, and a device for processing floating-point number. A self-defined floating-point number is obtained, wherein the self-defined floating-point number comprises a sign field, an exponent field, and a mantissa field, and a value of the self-defined floating-point number is determined by bit of the sign field, bits of the exponent field, bits of the mantissa field, and a bias value, wherein the bias value is determined by a total bit number of the exponent field. The self-defined floating-point number is applied to numerical calculations.