TensorLBM Platform

Dashboard

—

Total Jobs

—

Running

—

Completed

—

Failed

Platform Status

Loading…

Quick Start

Go to Pre-processing to generate geometry masks or convert physical units to LBM units.
Go to Solve, pick a simulation type, configure parameters, and click Submit Job.
Monitor progress in the Jobs sidebar (left).
Once complete, go to Post-process to view snapshots, plots, and metadata.
Use Benchmarks to run validation suites against published reference data.

Ship Hull CAD Modelling

Hull Parameters

Hull Type

Length L (lu)

Beam B (lu)

Draft T (lu)

Body-plan stations

Hull Form Coefficients

Click Preview to compute coefficients.

Ship → LBM Parameters

Ship length (m)

Speed (m/s)

ν (m²/s)

Froude target (0=off)

Hull Preview

Select hull type and parameters, then click Preview.

3D Hull Model (three.js)

3-D Voxel Mask

Generate the 3-D LBM obstacle mask for the CAD hull.

Launch LBM Solver (CAD → Fluid → Post-process)

Directly submit a 3-D ship-hull LBM simulation from the current CAD hull. The hull mask is passed to the solver; results appear in Post-process.

Hull length (lu)

Hull beam (lu)

Hull draft (lu)

Inlet velocity

Reynolds number

Smagorinsky Cs

Time steps

Output interval

Device

SUBOFF Submarine CAD Modelling

Submarine Parameters

Model Variant

Hull Length L (lu)

Max Radius R (lu, 0=auto)

Bow Fraction

Stern Fraction

Submarine Form Statistics

Click Preview to compute statistics.

Submarine Preview

Click Preview Submarine to generate.

Offshore Structure CAD

Structure Type

Diameter (lu)

Structure Statistics

Click Preview to compute statistics.

Offshore Structure Preview

Click Preview to generate.

Wageningen B-Series Propeller

Blade Parameters

Advance Ratio J

Pitch Ratio P/D

Blade Area Ratio Aₑ/A₀

Blades Z

Propeller Sizing

Thrust (N)

Va (m/s)

n (rev/s)

Open-Water Diagram

Select parameters and click Plot Curves.

Pre-processing

Polygon → Obstacle Mask (2D)

Grid width (nx)

Grid height (ny)

Vertices (one [x,y] per line, pixel coords)

Enter integer pixel coordinates, one pair per line.

Random Porosity Mask (2D)

Porosity (0–1)

Correlation length

Seed

LBM Unit Converter

Char. length L [m]

Char. velocity U [m/s]

Kinematic ν [m²/s]

LBM length (cells)

LBM velocity (lu)

Solver

Simulation Type

Configuration

Physical Model Layer Choose turbulence/multiphase/numerics before submitting.

Preset template

Flow type

Turbulence model

Multiphase model

Boundary condition

Numerical scheme

Smagorinsky C_s

Post-processing

Select Job

Summary

Select a job to view results.

No job selected.

Controls

Field

Colormap

Checkpoint

Velocity vectors

Streamlines

Field Viewer

Select a field and checkpoint, then click Render.

Select a job to view logs.

Select a job to view metadata.

Benchmarks

Marine & Ocean Engineering idle

Cylinder flow (St/Cd), sloshing tank (Faltinsen), near-bed pipeline (St), turbulent channel (log-law), Wigley hull (Cd), SUBOFF resistance (<3% error).

Mode

Device

Multiphase LBM idle

Static droplet (Laplace), spinodal decomposition, two-phase Poiseuille comparison (SCMC + CG).

Mode

Device

Lid-Driven Cavity – Ghia Comparison idle

Compare centreline velocity profiles against Ghia et al. (1982) reference data at Re = 100 / 400 / 1000.

Steps

Single-phase Accuracy Suite idle

Lid-driven cavity (Ghia), backward-facing step (Armaly), and rotating cylinder (Mittal) quantitative validation.

Mode

Device

MLUPS Performance idle

D2Q9 BGK throughput in Million Lattice Updates Per Second (MLUPS) at multiple grid sizes.

Grid sizes (comma-sep)

Steps

Porous Media idle

Laplace pressure, capillary invasion, and gas-water drainage benchmarks for porous-media multiphase flow.

Mode

Device

Compare runs

Select 2–10 completed jobs and view their derived parameters and diagnostics side by side.

Jobs

Loading…

AI Flow Transformer

Self-supervised training

nx ny Epochs Mask ratio

Deployment inference

Model ID (optional) Seed

Transformer models

Loading…

Result

HPC Orchestration + AI Governance

Experiment templates

Template Re sweep values (comma) nx ny n_steps output_interval

Submit result

Orchestration KPIs

Confidence gate

Prediction Baseline Uncertainty Max relative error Max uncertainty

Gate result

Active learning prioritization

top_k Candidates JSON

Prioritization result

AI Agent – Conversational modelling → compute → analysis

Loading agent info…

👋 你好！我是 TensorLBM 智能助手。我可以帮你完成"建模 → 求解 → 分析可视化"的全流程： - 请告诉我你想做的算例，例如：「跑一个 Re=200 的圆柱绕流，nx=200，迭代步数=2000」 - 也可以查询任务状态、分析结果、提取速度剖面等。 - Hi! I am the TensorLBM agent. Tell me a scenario (cylinder flow, dam break, ship hull, …) and I will submit it for you.

SUBOFF Surrogate Model

Data

Training

Epochs

Points

Device

Inference

N Points

Error Analysis

Visualization

Axis

Slice

Snap

Time-Series Animation

View

Start Snap

End Snap

Axis

Slice

Channels

Speed (fps)

Click Load to fetch animation frames. "Full tail field" shows the original LBM simulation data (200x50x200); "Training crop" shows the near-suboff region (100x50x100) used for AI training.

Status

Loading…

训练结果

#	Phase	Epoch	Loss(1e-4)	MSE(1e-4)	LR	Best Loss

推理结果

误差分析结果

Visualization 结果

Time-Series Animation 结果

General LBM Simulation

Geometry

Source

Physics

Density (kg/m³)

Viscosity (m²/s)

Inlet Velocity (m/s)

Reference Length (m)

Gravity [x,y,z]

Solver

Lattice

Collision

Resolution (cells/ref_length)

Domain Padding [xmin,xmax,ymin,ymax,zmin,zmax]

Device

Time Stepping

Max Steps

Warmup Steps

Snapshot Interval

Force Sample Interval

Smagorinsky Cs

Target Mach

Run

Contour Map

Force History

Cylinder Flow — Interactive Simulation

u_in

Radius

Steps/frame

Step 0 Not connected

Speed Field (drag cylinder to move)

Vorticity Field

Pressure (rho) Field

Speed Comparison — 某厂商 vs SDAA

u_in

Radius

Click Start to begin comparison

某厂商

Step 0 — ms/step

SDAA

Step 0 — ms/step