Multi-person online collaborative water and sediment numerical simulation method based on heterogeneous computing power of CPU and GPU

Abstract

The invention relates to the field of water and sediment numerical simulation, in particular to a multi-person online collaborative water and sediment numerical simulation method based on CPU and GPU heterogeneous computing power, and aims to save computing power resources, improve efficiency and realize computing power sharing. According to the technical scheme, the method comprises the steps that S1, an application is put forward, wherein a user prepares a water and sediment numerical simulation case and puts forward a computing power resource application to a water and sediment numerical simulation system; s2) program adaptation: if similar water and sediment numerical simulation cases exist in the water and sediment numerical simulation system, selecting an adapted water and sediment numerical simulation calculation program, otherwise, installing the water and sediment numerical simulation calculation program and realizing adaptation; s3) submitting a job: submitting the job by a user, and enabling the water and sediment numerical simulation cases to enter a corresponding calculation queue according to calculation power resources; s4) simulation calculation: a calculation unit of the water and sediment numerical simulation system performs calculation according to the queue, and a calculation result is stored in a file unit of the water and sediment numerical simulation system; and S5) result analysis: extracting a calculation result by a user, and carrying out water and sediment numerical simulation application.

Description

Technical Field

[0001] The present invention relates to the field of water and sediment numerical simulation, and specifically to a method for multi-person online collaborative water and sediment numerical simulation based on heterogeneous computing power of CPU and GPU. Background Art

[0002] Water and sediment numerical simulation provides important scientific basis and technical support for estuary regulation by providing high-precision simulation results and scientific prediction and analysis, and is of great significance for solving practical problems such as estuary siltation, erosion, and typhoon storm surge.

[0003] In recent years, parallel computing has been widely applied to water and sediment numerical simulation. For example, Xia Qing from Tsinghua University applied cluster computing method to carry out basin hydrological process simulation, and Wang Jianjun et al. from the Key Laboratory of Engineering Sediment and Transportation Industry of Tianjin Research Institute for Water Transport Engineering, Ministry of Transport studied the parallel computing technology of two-dimensional water and sediment mathematical model of river channels and successfully developed the inland water and sediment numerical simulation software TK-2DC. The invention patent "Multi-scale hydrodynamic coupling method based on FVCOM and OpenFOAM models" studied the multi-scale hydrodynamic coupling method.

[0004] Through research, the main problem existing in the current water and sediment numerical simulation is that the calculation program runs independently on the physical server, using only CPU computing power or GPU computing power. This exclusive use of high-performance computing cluster resources not only wastes computing power resources but also cannot achieve computing power sharing, making it difficult to form economies of scale. Summary of the Invention

[0005] The purpose of the present invention is to overcome the deficiencies in the above background art and provide a method for multi-person online collaborative water and sediment numerical simulation based on heterogeneous computing power of CPU and GPU, so as to save computing power resources, improve efficiency, and achieve computing power sharing.

[0006] The technical solution of the present invention is as follows:

[0007] A method for multi-person online collaborative water and sediment numerical simulation based on heterogeneous computing power of CPU and GPU includes the following steps:

[0008] S1) Submitting an application: The user prepares a water and sediment numerical simulation case and submits an application for computing power resources to the water and sediment numerical simulation system;

[0009] S2) Adapting the program: If there is a similar water and sediment numerical simulation case in the water and sediment numerical simulation system, select the already adapted water and sediment numerical simulation calculation program; otherwise, install the water and sediment numerical simulation calculation program and achieve adaptation;

[0010] S3) Submitting a job: The user submits a job, and the water and sediment numerical simulation case enters the corresponding calculation queue according to the computing power resources;

[0011] S4) Simulation calculation: The calculation units of the water and sediment numerical simulation system perform calculations according to a queue, and the calculation results are stored in the file unit of the water and sediment numerical simulation system;

[0012] S5) Result analysis: The user extracts the calculation results and conducts water and sediment numerical simulation applications.

[0013] The water and sediment numerical simulation system includes a calculation unit, a network unit, and a file unit.

[0014] The water and sediment numerical simulation calculation program includes: 1) Designing the water and sediment numerical simulation calculation equation; 2) Discretizing the calculation equation, writing code using the Fortran language, and compiling it into an executable file; 3) Determining the area and time period to be simulated; 4) Calibrating sediment parameters using measured data.

[0015] The calculation unit includes several CPU calculation nodes, several GPU calculation nodes, and several cluster nodes; the network unit includes a BMC network switch, a service network switch, a first IB network switch, and a second IB network switch; the file unit includes several storage nodes.

[0016] The BMC network switch and the service network switch connect the calculation unit and the file unit; the first IB network switch connects the second IB network switch, the calculation unit, and the file unit; the second IB network switch connects the calculation unit.

[0017] In the step S1), the water and sediment numerical simulation case includes establishing the research area, boundary conditions, and initializing model parameters of the water and sediment numerical simulation case; the computing power resources include the computing power of CPU calculation nodes, the computing power of GPU calculation nodes, and the computing power of CPU and GPU calculation nodes.

[0018] In the step S3), the calculation queue includes a CPU calculation node queue, a GPU calculation node queue, and a CPU and GPU calculation node queue; each calculation queue contains several calculation tasks. The user can adjust the order through strategies, start, pause, and end the calculation tasks; the current calculation task starts the water and sediment numerical simulation calculation program of the calculation unit through a command. After the calculation unit completes the calculation, it is automatically cleared from the current calculation node queue.

[0019] The step S5) includes: The user extracts and analyzes the calculation results. If the calculation results meet the requirements, water and sediment numerical simulation applications are carried out. Otherwise, the user modifies the initial model parameters of the water and sediment numerical simulation case, or modifies and then installs the water and sediment numerical simulation calculation program first, and enters step S3).

[0020] The design of the water and sediment numerical simulation calculation equation includes:

[0021] 1.1) Basic calculation formulas for designing a three-dimensional hydrodynamic model

[0022] In the three-dimensional hydrodynamic model, the sigma coordinate that fits the water depth is adopted vertically, and its definition is:

[0023]

[0024] where: z is the actual water depth, H is the long-term average water depth; η is the sea surface undulation; D is the true water depth, D = H + η; The value range is [-1, 0], -1 at the bottom of the water, and 0 at the free sea surface.

[0025] In three-dimensional space, without considering the Reynolds average and hydrostatic assumptions under wave action for the time being, the continuity equation, horizontal momentum equation, and temperature T and salinity s equations are respectively:

[0026]

[0027]

[0028] where:

[0029] T is time; x α and x β , are horizontal coordinates; ω is the vertical velocity under the vertical sigma coordinate; g is the acceleration due to gravity; F f is the Coriolis force vector; F h , F T and F s are respectively the horizontal momentum mixing term and the horizontal temperature and salinity mixing terms; K m and K v are respectively the vertical momentum eddy viscosity coefficient and the vertical temperature and salinity eddy viscosity coefficient.

[0030] 1.2) Design of the bottom and surface boundary conditions of the three-dimensional hydrodynamic model

[0031] The surface and bottom boundary conditions of the flow velocity are respectively:

[0032]

[0033] where: and are respectively the sea surface wind stress and the bottom friction stress, ρ0 is the density of fresh water under normal temperature and pressure;

[0034] The surface and bottom boundary conditions of the temperature are respectively:

[0035]

[0036] where: Q nTo represent the net heat flux; SW is the short-wave radiation reflected by the sea surface; C p is the specific heat of seawater; A H is the horizontal mixing coefficient of temperature; λ is the terrain slope, and n is the horizontal coordinate;

[0037] The sea surface and bottom boundary conditions of salinity are as follows:

[0038]

[0039] Among them: and are the evaporation and precipitation rates respectively.

[0040] 1.3) Design a sediment dynamics model

[0041] When not considering the horizontal diffusion of suspended sediment, the control equation of suspended sediment is:

[0042]

[0043] Among them: C is the concentration of suspended sediment in the water body; w s is the settling velocity of suspended sediment, and K v is the vertical eddy viscosity coefficient of suspended sediment;

[0044]

[0045] Among them: C d is the bottom friction coefficient; κ is the von Kármán constant; z b is the thickness of the near-bottom water body; z 0b is the characteristic value of bottom roughness; A = 5.5 is an empirical constant;

[0046] The flux Richardson number R f is:

[0047] R f = 0.725[R i + 0.186 - (R i 2 - 0.316R i + 0.0346)] 12 (14)

[0048] R i is the Richardson number;

[0049] The influence of sediment concentration on the state equation is:

[0050]

[0051] Among them: ρ w is the seawater density; ρ s is the volume density of sediment.

[0052] 1.4) Design the sedimentation and resuspension processes of suspended sediment

[0053] There is no vertical flux of suspended sediment at the free sea surface. Therefore, the sea surface boundary condition is as follows:

[0054]

[0055] The net sediment flux E of suspended sediment perpendicular to the seabed b is the difference between the sedimentation flux and the resuspension flux:

[0056]

[0057] where: E b is the net sediment flux; E0 is the erosion rate constant, C b is the near-bottom suspended sediment concentration; τ b is the bottom shear stress; τ ce and τ cd are the critical shear stress and the critical sedimentation stress, respectively;

[0058] The sedimentation velocity expression is as follows:

[0059]

[0060] where: w s0 is the free sedimentation velocity, C0 is the critical suspended sediment concentration for flocculation to occur, and m1, m2, n1, and n2 are empirical parameters of the sedimentation velocity.

[0061] The determination of the area and time period to be simulated includes: using SMS software to design an unstructured computational grid, taking Matlab as the platform, calling the t_tide function, and obtaining the open boundary tide level fluctuation sequence corresponding to the time period as the open boundary forcing input into the program.

[0062] The calibration of sediment parameters includes:

[0063] Calibrate the sediment parameters in step 4), and by adjusting the parameters therein, make the SS parameter greater than 0.65;

[0064]

[0065] where: SS is the skill score, X model is the output result of the numerical model, X observation is the field observation result, and X observation is the average of the observation results over time.

[0066] The beneficial effects of the present invention are:

[0067] The present invention adopts heterogeneous computing power combining CPU and GPU, optimizes the water and sediment numerical simulation algorithm, realizes multi-user online collaborative applications, improves the sharing efficiency of computing power resources, and solves the problems of difficult scheduling, low efficiency, and much waste of computing power resources in the high-performance computing cluster for water and sediment numerical simulation. BRIEF DESCRIPTION OF THE DRAWINGS

[0068] Figure 1 is a flowchart of the present invention.

[0069] Figure 2 is a schematic diagram of the water and sediment numerical simulation system of the present invention. DETAILED DESCRIPTION OF THE INVENTION

[0070] In order to make the objectives, technical solutions and advantages of the present invention clearer, the present invention will be further described in detail below with reference to the accompanying drawings and embodiments. It should be understood that the specific embodiments described herein are only for explaining the present invention and are not intended to limit the present invention.

[0071] As Figure 1 shown, a method for multi-person online collaborative water and sediment numerical simulation based on heterogeneous computing power of CPU and GPU includes the following steps:

[0072] S1) Submitting an application

[0073] The user prepares a water and sediment numerical simulation case and submits an application for computing power resources to the water and sediment numerical simulation system (logging in to the service network server). After passing the review, it enters step S2).

[0074] The water and sediment numerical simulation case includes establishing the research area, boundary conditions, and initializing model parameters of the water and sediment numerical simulation case.

[0075] The computing power resources include CPU computing node computing power, GPU computing node computing power, and CPU and GPU computing node computing power.

[0076] The water and sediment numerical simulation system includes a computing unit, a network unit, and a file unit.

[0077] S2) Adapting the program

[0078] If there is a similar water and sediment numerical simulation case in the water and sediment numerical simulation system, select the already adapted water and sediment numerical simulation calculation program. Otherwise, install the water and sediment numerical simulation calculation program and implement the adaptation. Different computing power resources need to adapt to the corresponding water and sediment numerical simulation calculation program.

[0079] The water and sediment numerical simulation calculation program includes: 1) designing the water and sediment numerical simulation calculation equation; 2) discretizing the calculation equation, writing code using Fortran language, and compiling it into an executable file; 3) determining the area and time period to be simulated; 4) calibrating sediment parameters using measured data.

[0080] S3) Submitting a job

[0081] The user submits a job, and the water and sediment numerical simulation case enters the corresponding calculation queue according to the selected computing resources. After the job is submitted, the user can view the job running and resource usage status at any time.

[0082] In the step S3, the calculation queue includes a CPU computing node queue, a GPU computing node queue, and a CPU and GPU computing node queue.

[0083] Each calculation queue contains several calculation tasks. The user can adjust the order through strategies and start, pause, or end the calculation tasks; the current calculation task starts the water and sediment numerical simulation calculation program of the computing unit through a command. After the computing unit completes the calculation, it is automatically cleared from the current computing node queue.

[0084] S4) Simulation calculation

[0085] The computing unit of the water and sediment numerical simulation system calculates according to the queue, and the calculation results are stored in the file unit of the water and sediment numerical simulation system.

[0086] S5) Result analysis

[0087] The user extracts the calculation results and conducts analysis. If the calculation results meet the requirements, the user uses an external system to carry out water and sediment numerical simulation applications. Otherwise, the user modifies the initial model parameters of the water and sediment numerical simulation case or reinstalls the water and sediment numerical simulation calculation program and enters step S3).

[0088] Due to the limited storage capacity of the file unit, the user must process all data in a timely manner.

[0089] The following further explains the water and sediment numerical simulation system.

[0090] The computing unit includes several CPU computing nodes, several GPU computing nodes, and several cluster nodes. The CPU computing node includes an Intel XEON Platinum 8358 processor. The GPU computing node includes an A800 graphics card.

[0091] The network unit includes a BMC network switch, a service network switch, a first IB network switch, and a second IB network switch. The file unit includes several storage nodes.

[0092] The BMC network switch is connected to the service network switch, the computing unit, and the file unit; the first IB network switch is connected to the second IB network switch, the computing unit, and the file unit; the second IB network switch is connected to the computing unit.

[0093] The BMC network switch includes a 1 Gb / s electrical port network switch. The service network switch includes a 40 Gb / s Ethernet optical port network switch. The first IB network switch includes a 200 Gb / s Infiniband network switch. The second IB network switch includes a 400 Gb / s Infiniband network switch.

[0094] The storage node includes a hybrid distribution of SSD hard drives and HDD hard drives for storing relevant files, which can be expanded as needed.

[0095] The BMC network switch is connected to the CPU computing node and the cluster node. Through the cluster node, it provides functions such as installing the operating system on the CPU node, modifying device parameters, upgrading plugins, cluster control, and maintaining operation logs; the BMC network switch is connected to the GPU computing node and the cluster node. Through the cluster node, it provides functions such as installing the operating system on the GPU node, modifying device parameters, upgrading plugins, cluster control, and maintaining operation logs; the BMC network switch is connected to the storage node to provide file access functions.

[0096] The service network switch is connected to the CPU computing node and the cluster node. Through the cluster node, it provides functions such as CPU computing power resource application services and collaborative services; the service network switch is connected to the GPU computing node and the cluster node. Through the cluster node, it provides functions such as GPU computing power resource application services and collaborative services; the service network switch is connected to the storage node to provide file access functions.

[0097] The first IB network switch is connected to the CPU computing node and the storage node to provide high-speed file access functions for the CPU computing node; the first IB network switch is connected to the GPU computing node and the storage node to provide high-speed file access functions for the GPU computing node; the first IB network switch is connected to the cluster node and the storage node to provide high-speed file access functions for the cluster node; the first IB network switch is connected to the second IB network switch to provide interface conversion functions.

[0098] The second IB network switch is connected to the CPU computing node and the cluster node to provide internal network communication and file exchange functions for CPU parallel computing; the second IB network switch is connected to the GPU computing node and the cluster node to provide internal network communication and file exchange functions for GPU parallel computing.

[0099] The system is configured with 2 cluster nodes, 10 CPU computing nodes, and 2 GPU computing nodes.

[0100] The configuration of the CPU computing nodes is as follows: CPU: Xeon 8358 8*2, Memory: DDR4 4800 32G*16, Hard Disk: 960G 2.5-inch SATA 6Gb SSD*2, 8i SAS array card*1, Network Card: MCX75310AAS-NEAT 400G NDR IB network card*1, Power Supply: Configured with (N+N) redundant power supply.

[0101] The configuration of the GPU computing nodes is as follows: CPU: AMD EPYC 9654*2, Memory: DDR5 4800MHz 64G*24, Hard Disk: 960GB Intel / Samsung data center-class SSD*2, 2G 8i RAID card*1, Network Card: NVIDIA HGX NVLINK H800 8-GPU 80GB, MCX75310AAS-NEAT 400G NDR IB network card*4, 4-port gigabit network card*1, Power Supply: Configured with (N+N) redundant power supply.

[0102] The configuration of the cluster nodes is as follows: CPU: Xeon 4310*2, Memory: DDR4 32G*4, Disk Array: 4GB SAS 8-port RAID card, Hard Disk: 600G 2.5-inch 10k SAS 12Gb hard disk*3, 8-drive bay backplane, Network Card: 4-port gigabit electrical network card*1, Single-port 200G HDR IB card, Power Supply: Configured with (N+N) redundant power supply.

[0103] The configuration of the BMC network switch and the service network switch is as follows: 48 electrical ports, 4 ten-gigabit SFP+ optical ports, fully configured with 40Gb / s multimode optical modules.

[0104] The configuration of the first IB network switch and the second IB network switch is as follows: 64-port HDR IB switch.

[0105] The configuration of the storage node is as follows:

[0106] (1) Performance layer: CPU: 5318Y*2, Memory: 256G, System Disk: 480G SATA SSD*2, Data Disk: 3.84T NVME SSD PCIe 4.0*24, Network Card: 200G HDR IB*2, 10G (optical)*2, 4-port gigabit network card*1, Power Supply: Configured with (N+N) redundant power supply.

[0107] (2) Capacity layer: CPU: 4314 * 2, Memory: 128GB, System disk: 480GB SATA SSD * 2, Log disk: 7.68TB NVMe * 4, Data disk: 10TB NL - SAS HDD * 36, RAID card: 9440 * 1, Network card: 200G HDR IB * 1, 10G (optical) * 2, 4 - port Gigabit network card * 1, Power supply: Configure (N + N) redundant power supply.

[0108] The following further explains the water - sediment numerical simulation calculation program.

[0109] 1) Design the water - sediment numerical simulation calculation equation

[0110] 1.1) Design the basic calculation formulas required for the three - dimensional hydrodynamic model

[0111] In the three - dimensional hydrodynamic model, the sigma coordinate fitting the water depth is adopted vertically, and its definition is:

[0112]

[0113] In formula (1): z is the actual water depth, H is the long - term average water depth; η is the sea - surface undulation; D is the true water depth, D = H + η, The value range is [-1, 0], -1 at the bottom of the water, and 0 at the free sea surface;

[0114] In three - dimensional space, without considering the Reynolds average and hydrostatic assumptions under wave action for the time being, the continuity equation, horizontal momentum equation, and temperature T, salinity s equations can be written respectively as:

[0115]

[0116] In formulas (2), (3), (4), (5):

[0117] T is time, x and y represent the eastward and northward directions in Cartesian coordinates respectively; in order to simplify the formula expression, α and β in the formulas include both the x and y directions. For example, in formula (3), when α represents the x direction, at this time U α is the velocity in the east - west direction, and the eastward direction is positive. At this time, U β is the velocity in the north - south direction, and the northward direction is positive; when α represents the y direction, at this time U α is the velocity in the north - south direction, and the northward direction is positive. At this time, U β is the velocity in the east - west direction, and the eastward direction is positive; the horizontal coordinates are x α and x β , corresponding to x and y;

[0118] ω is the vertical velocity in the vertical sigma coordinate; g is the acceleration due to gravity; F fis the Coriolis force vector (-fv, fu); F h , F T and F s are the horizontal momentum mixing term and the horizontal temperature and salinity mixing terms calculated through the Smagorinsky eddy parameterization scheme (Smagorinsky, 1963);

[0119] K m and K v are the vertical momentum eddy viscosity coefficient and the vertical temperature and salinity eddy viscosity coefficient calculated through the Mellor - Yamada level 2.5 turbulence closure model (Mellor and Yamada, 1982).

[0120] 1.2) Design the bottom boundary conditions of the three - dimensional hydrodynamic model

[0121] The sea - surface and bottom boundary conditions of the flow velocity (without considering evaporation, precipitation, and groundwater) are respectively:

[0122]

[0123] In equations (6) and (7): and are the sea - surface wind stress and the bottom friction stress respectively, ρ0 is the density of fresh water under normal temperature and pressure; The algorithm of C d is summarized in step 3);

[0124] The sea - surface and bottom boundary conditions of the temperature are respectively:

[0125]

[0126] In equations (8) and (9), Q n is the net surface heat flux, which is the sum of the downward net short - wave radiation, the downward net long - wave radiation, the latent heat, and the sensible heat; SW is the short - wave radiation reflected by the sea surface; C p is the specific heat of seawater; A H is the horizontal mixing coefficient of temperature, calculated from the Smagorinsky eddy parameterization scheme (Smagorinsky, 1963); λ is the terrain slope, and n is the horizontal coordinate.

[0127] The sea - surface and bottom boundary conditions of the salinity are respectively:

[0128]

[0129] In equations (10) and (11), and are the evaporation and precipitation rates respectively.

[0130] 1.3) Design the sediment dynamics model

[0131] The suspended sediment control equation when not considering the horizontal diffusion of suspended sediment is as follows:

[0132]

[0133] In Equation (12): C is the concentration of suspended sediment in water (unit: kg·m 3 or g·L), w s is the settling velocity of suspended sediment, and K v is the vertical eddy viscosity coefficient of suspended sediment.

[0134] Taking the Oujiang Estuary as an example: Considering that the Oujiang Estuary is located in the high-turbidity sea area along the coast of Zhejiang and Fujian in China, the sediment concentration in the maximum turbidity zone often exceeds 1 kg·m -3 . In this case, the stratification caused by sediment in the bottom boundary layer will lead to a weakening of the vertical eddy viscosity coefficient, thereby reducing the bottom friction. Wang et al. (2002) incorporated the influence of the bottom boundary layer stratification caused by suspended sediment on the bottom shear stress through the flux Richardson number R f into the calculation of the bottom friction coefficient C d :

[0135]

[0136] In Equation (13): κ is the von Kármán constant; z b is the thickness of the near-bottom water body; z 0b is the characteristic value of the bottom roughness; A = 5.5 is an empirical constant.

[0137] The flux Richardson number R f is a formula used to characterize the vertical density stratification in the Mellor-Yamada 2.5 turbulence closure model, and the calculation method is as follows (Mellor and Yamada, 1974):

[0138] R f = 0.725[R i + 0.186 - (R i 2 - 0.316R i + 0.0346) 12 (14)

[0139] In addition, in order to represent the influence of the suspended sediment concentration on the hydrodynamic force, R i is the Richardson number, and the influence of the sediment concentration on the state equation is written as:

[0140]

[0141] In Equation (15), ρ w is the seawater density; ρ sVolume density of sediment.

[0142] 1.4) Design the sedimentation and resuspension processes of suspended sediment

[0143] There is no vertical flux of suspended sediment at the free sea surface. Therefore, the sea surface boundary condition is:

[0144]

[0145] The net sediment flux E of suspended sediment perpendicular to the seabed b is the difference between the sedimentation flux and the resuspension flux:

[0146]

[0147] In equations (16) and (17), the net sediment flux E b is usually defined as (Ariathurai and Krone, 1976):

[0148]

[0149] In equation (18): E0 is the erosion rate constant, C b is the near-bottom suspended sediment concentration; τ b is the bottom shear stress; τ ce and τ cd are the critical shear stress and the critical sedimentation stress, respectively;

[0150] For fine-grained sediment, the role of sediment flocculation process in the sedimentation velocity w s becomes non-negligible. Therefore, the sedimentation velocity expression including the three sedimentation stage processes of "free sedimentation", "flocculent sedimentation" and "hindered sedimentation" can be written as (Mehta and McAnally, 2008):

[0151]

[0152] In equation (19): w s0 is the free sedimentation velocity, C0 is the critical suspended sediment concentration for flocculation to occur, and m1, m2, n1 and n2 are the empirical parameters of the sedimentation velocity.

[0153] 2) Discrete calculation equations, write code using Fortran language and compile it into an executable file.

[0154] Based on the discrete method of finite volume, where the flow velocity U αIt is defined at the centroid of each triangle and is called the "triangle element", while other scalar variables are defined at the vertices of the triangle and are called "nodes". To improve the computational efficiency, the model adopts the "separation algorithm of inner and outer models", in which the calculations of slow motion processes such as three-dimensional momentum, temperature, salinity, suspended sediment and waves are set with a longer inner model time step, while the faster-changing two-dimensional water level change uses a shorter outer model time step.

[0155] 3) Determine the area and time period to be simulated

[0156] Use software such as SMS to design an unstructured computational grid. Taking Matlab as the platform, call the t_tide function to obtain the open boundary tide level fluctuation sequence corresponding to the time period, and input it into the program as the open boundary forcing.

[0157] 4) Calibrate the sediment parameters using measured data

[0158] Use various survey instruments and equipment for offshore observations to obtain measured data such as hourly flow velocity, flow direction, salinity and sediment concentration to calibrate the sediment parameters in step S4). By adjusting the parameters, make the SS parameter greater than 0.65;

[0159] Where: SS is the "Skill Score" (SS), which is the root mean square error (RMSE) between the standardized observed data and the model results, indicating the accuracy of the numerical model simulation. The closer it is to "1", the higher the accuracy of the simulation model (Murphy, 1988):

[0160]

[0161] In Equation (20), X model is the output result of the numerical model, X observation is the field observation result, is the average of the observation results over time.

[0162] Preferred embodiments of the present invention are shown in the accompanying drawings. However, the present invention can be implemented in many different forms and is not limited to the embodiments described in this specification. On the contrary, the purpose of providing these embodiments is to make the understanding of the disclosure of the present invention more thorough and comprehensive.

Claims

1. A multi-person online collaborative water and sand numerical simulation method based on CPU and GPU heterogeneous computing power, comprising the following steps: S1) Submit application: The user prepares a water-sediment numerical simulation case and submits an application for computing resources to the water-sediment numerical simulation system; S2) Adaptation program: if the water-sand numerical simulation system has similar water-sand numerical simulation cases, select the adapted water-sand numerical simulation calculation program; otherwise, install the water-sand numerical simulation calculation program and implement the adaptation; S3) Submitting a job: The user submits a job, and the water-sand numerical simulation case enters the corresponding computing queue according to the computing resources; S4) simulation calculation: the calculation unit of the water-sand numerical simulation system performs calculation according to the queue, and the calculation result is stored in the file unit of the water-sand numerical simulation system; S5) Results analysis: Users extract calculation results and carry out water and sediment numerical simulation applications; The water-sand numerical simulation system comprises a computing unit, a network unit, and a file unit; The water-sand numerical simulation calculation program includes: 1) designing water-sand numerical simulation calculation equations; 2) discretizing the calculation equations, writing codes using the Fortran language, and compiling them into executable files; 3) determining the area and time period to be simulated; 4) calibrating sediment parameters using measured data.

2. According to claim 1, a multi-person online collaborative water and sand numerical simulation method based on heterogeneous computing power of CPU and GPU is characterized by: The computing unit includes several CPU computing nodes, several GPU computing nodes, and several cluster nodes; the network unit includes a BMC network switch, a business network switch, a first IB network switch, and a second IB network switch; and the file unit includes several storage nodes.

3. The method for numerical simulation of water and sand based on multi-person online collaboration of CPU and GPU heterogeneous computing power according to claim 2 is characterized by: The BMC network switch and the business network switch connect the computing unit and the file unit; the first IB network switch connects the second IB network switch, the computing unit and the file unit; the second IB network switch connects the computing unit.

4. The method for numerical simulation of water and sand based on multi-person online collaboration of CPU and GPU heterogeneous computing power according to claim 1 is characterized by: In the step S1), the water-sand numerical simulation case includes establishing a research area, boundary conditions, and initialization model parameters for the water-sand numerical simulation case; the computing power resources include CPU computing node computing power, GPU computing node computing power, and CPU and GPU computing node computing power.

5. The method for numerical simulation of water and sand based on multi-person online collaboration of CPU and GPU heterogeneous computing power according to claim 1 is characterized by: In the step S3), the computing queue includes a CPU computing node queue, a GPU computing node queue, and a CPU and GPU computing node queue; Each computing queue contains several computing tasks. Users can start, pause, and end computing tasks by adjusting the order of priority through strategies. The current computing task starts the water and sand numerical simulation calculation program of the computing unit through commands. After the computing unit completes the calculation, it is automatically cleared from the current computing node queue.

6. The method for numerical simulation of water and sand based on multi-person online collaboration of CPU and GPU heterogeneous computing power according to claim 1, characterized in that: The step S5) includes: the user extracts the calculation results and analyzes them. If the calculation results meet the requirements, the water-sand numerical simulation application is carried out. Otherwise, the user modifies the initialization model parameters of the water-sand numerical simulation case, or modifies and then installs the water-sand numerical simulation calculation program, and enters step S3).

7. The method for numerical simulation of water and sand based on multi-person online collaboration of CPU and GPU heterogeneous computing power according to claim 1 is characterized by: The design water and sand numerical simulation calculation equation includes: 1.1) Basic calculation formulas required for designing a three-dimensional hydrodynamic model The sigma coordinate of the fitted water depth is used in the vertical direction of the three-dimensional hydrodynamic model, which is defined as: Where: z is the actual water depth, H is the long-term average water depth; η is the sea surface fluctuation; D is the true water depth, D = H + η; The value range is [-1,0], where the bottom of the water is -1 and the free sea surface is 0; In three-dimensional space, without considering the Reynolds average and static pressure assumptions under the action of waves, the continuity equation, horizontal momentum equation, temperature T, and salinity s equation are: in: T is time; x α and x β , is the horizontal coordinate; ω is the vertical velocity under the vertical sigma coordinate; g is the gravitational acceleration; F f is the Coriolis force vector; F h 、F T and F s are the horizontal momentum mixing term and the horizontal temperature and salt mixing term respectively; K m and K v are the vertical momentum eddy viscosity coefficient and the vertical temperature-salinity eddy viscosity coefficient respectively; 1.2) Design of surface and bottom boundary conditions for the three-dimensional hydrodynamic model The surface and bottom boundary conditions of the flow velocity are: Where: τ sα and τ bα are the sea surface wind stress and the seabed friction stress, ρ0 is the density of fresh water at normal temperature and pressure; The sea surface and bottom boundary conditions of temperature are: Where: Q n is the net heat flux; SW is the shortwave radiation reflected by the sea surface; C p is the specific heat of seawater; A H is the horizontal mixing coefficient of temperature; λ is the terrain slope, and n is the horizontal coordinate; The sea surface and bottom boundary conditions of salinity are: in: and are the evaporation and precipitation rates, respectively; 1.3) Design sediment dynamics model The suspended sediment control equation when the horizontal diffusion of suspended sediment is not considered is: Where: C is the concentration of suspended sediment in the water; w s is the settling velocity of suspended sediment, K v is the vertical eddy viscosity coefficient of suspended sediment; Where: C d is the bottom friction coefficient; κ is the von Karman constant; z b is the thickness of the water body near the bottom; 0b is the characteristic value of bottom roughness; A=5.5 is the empirical constant; Flux Richardson number R f for: R f =0.725[R i +0.186-(R i 2 -0.316R i +0.0346) 1 / 2 ] (14) R i is the Richardson number; The effect of sediment concentration on the state equation is: Where: w is the density of seawater; ρ s Bulk density of sediment; 1.4) Design the sedimentation and resuspension process of suspended sediment There is no vertical flux of suspended sediment on the free sea surface, so the sea surface boundary condition is: The net sediment flux E of suspended sediment vertically to the seafloor b is the difference between the sedimentation flux and the resuspension flux: Where: E b is the net sediment flux; E0 is the erosion rate constant, C b is the concentration of suspended sediment near the bottom; τ b is the bottom shear stress; τ ce and τ cd are the critical shear stress and critical settlement stress, respectively; The sedimentation velocity expression is: Where: w s0 is the free settling velocity, C0 is the critical suspended sediment concentration for flocculation to occur, and m1, m2, n1 and n2 are empirical parameters of the settling velocity.

8. The method for numerical simulation of water and sand based on multi-person online collaboration of CPU and GPU heterogeneous computing power according to claim 1 is characterized by: The determination of the area and time period to be simulated includes: using SMS software to design an unstructured computational grid, using Matlab as a platform, calling the t_tide function, and reporting an open boundary tidal level fluctuation sequence of the corresponding time period as an open boundary forced input into the program.

9. The method for numerical simulation of water and sand based on multi-person online collaboration of CPU and GPU heterogeneous computing power according to claim 1, characterized in that: The rated sediment parameters include: Calibrate the sediment parameters in step 4) by adjusting the parameters so that the SS parameter is greater than 0.65; Among them: SS is the skill score, X model is the output result of the numerical model, X observation The results of on-site observations are is the average of the observations over time.

Images