云熔智算 · v0.9 BETA 由新加坡国立大学闫文韬教授团队研发 Developed by Prof. Wentao Yan's team at NUS

首印即成
非凡智造
First time right
3D printing

全球领先的 3D 打印仿真平台。打印前精准预测结果、优化参数,大幅提升打印成功率、减少试错次数、缩短工艺开发周期。 The world's leading 3D printing simulation platform. Predict results and optimize parameters before you print — boosting success rates, cutting trial iterations, and shortening process development cycles.

选择性激光熔化 SLMSLM· 电子束熔化 EBMEBM· 粘结剂喷射BINDER JET· Ti-6Al-4V· 316L· INCONEL 718· AlSi10Mg· 高熵合金HIGH-ENTROPY ALLOYS· TIB₂/AL MMC· CoCr· 选择性激光熔化 SLMSLM· 电子束熔化 EBMEBM· 粘结剂喷射BINDER JET· Ti-6Al-4V· 316L· INCONEL 718
仿真精度SIM ACCURACY
82.3%
vs 商业软件 62.2%
vs commercial 62.2%
AI 推理AI INFERENCE
10ms
对比 CPU 2小时44分
vs 2h 44m on CPU
GPU 加速GPU SPEEDUP
10×
GPU 并行 vs 单线程
parallel vs single-thread
仿真功能SIM FUNCTIONS
40+
10+ 独有功能
10+ proprietary
真实仿真画面 · 自动播放 Live simulation footage · auto-playing

看着仿真
从一颗粉末跑到整个零件
Watch a simulation grow
from a single grain to the full part

Thermal stress evolution
Mechanical properties — strain animation Deformed mesh — final state
SIM · LIVE
01 / 04
粉末PARTICLE316L · 30 μm
熔池温度MELT TEMP~1800 K
扫描速度SCAN VEL1.2 m/s
晶粒模型GRAIN IDCA + PHASE FIELD
尺度SCALE50 μm
最大应力 σσ MAX3.2 GPa
尺度SCALE零件 · 25 mmPART · 25 mm
延度DUCTILITYε 24%
疲劳寿命FATIGUE10⁶ 次循环10⁶ CYCLES
阶段 01 / 05STAGE 01 / 05

粉末铺设

Powder spreading

颗粒尺度的离散元仿真重现真实粉末床 — 摩擦、范德华力、颗粒分级一颗不漏。这是后续所有物理的起点。

Particle-scale discrete-element simulation replicates real powder beds — friction, van der Waals, size segregation, every grain modeled. This is where every later physics step starts.

  • 铺粉均匀度预测
  • Spreading uniformity
  • 颗粒分级 / 卷起
  • Particle segregation / pickup
  • 粘结剂喷射耦合
  • Binder jetting coupling
阶段 02 / 05STAGE 02 / 05

熔池动力学

Melt-pool dynamics

激光扫过的瞬间发生了什么?热流、马兰戈尼对流、蒸发反冲压、匙孔失稳 — 多物理场耦合实时求解,温度场以颜色直接显示。

What happens the instant the laser passes? Heat flow, Marangoni convection, evaporation recoil, keyhole instability — all multi-physics fields coupled and solved in real time. Color encodes temperature.

  • Ray-tracing 真实热源
  • Ray-traced heat source
  • 蒸发 / 匙孔效应
  • Evaporation / keyholing
  • 磁场 · 超声 · 极端环境
  • Magnetic · ultrasound · extreme env.
阶段 03 / 05STAGE 03 / 05

微观结构演化

Microstructure evolution

熔池凝固时晶粒是怎么长出来的?相场建模逐层模拟形核、长大、粗化 — 不同晶粒方向用不同颜色,让"看不见的合金组织"变成可视可量化的数据。

How do grains grow as the melt pool solidifies? Phase-field modeling traces nucleation, growth and coarsening layer-by-layer — each grain orientation a distinct color, turning invisible microstructure into quantitative data.

  • 相场晶粒建模
  • Phase-field grain model
  • 枝晶 / 收缩孔识别
  • Dendrites / shrinkage pores
  • 抗热裂判据
  • Hot-crack criterion
阶段 04 / 05STAGE 04 / 05

热应力 / 残余应力

Thermal & residual stress

层层堆积后,残余应力如何分布?一个 3D 色谱清晰展示从蓝色(拉应力 −130 MPa)到红色(压应力 3.2 GPa)的真实分布 — 直接告诉你哪儿会翘曲、哪儿会裂。

After layers stack up, how is residual stress distributed? A 3D color map shows the real distribution from blue (tension −130 MPa) to red (compression 3.2 GPa) — telling you exactly where warpage and cracking will appear.

  • 晶粒位错密度
  • Dislocation density
  • 整件翘曲预测
  • Bulk warpage prediction
  • 裂纹萌生位置
  • Crack initiation sites
阶段 05 / 05STAGE 05 / 05

机械性能预测

Mechanical properties

从微观尺度的应力分布出发,外推到宏观零件的延展性、疲劳寿命与断裂行为 — 让工程师在打印之前就知道这个零件能不能用、能用多久。

From microscopic stress distributions out to bulk ductility, fatigue life and fracture behavior — engineers know if the part will work, and how long it will last, before it's ever printed.

  • 延展性预测
  • Ductility prediction
  • 疲劳寿命估算
  • Fatigue life estimation
  • 断裂力学分析
  • Fracture mechanics

仿真不是目的 Simulation isn't the goal. 首印即成才是 First-print success is.

01

提升打印成功率 Higher print success rate

打印前在虚拟环境中完整模拟过程,预测缺陷与形变。金属打印 20–30% 的失败率,变成一次成型的把握。

Run the full print in a virtual environment first — defects and distortion predicted before you commit material. Turn 20–30% metal-AM failure into first-print confidence.

打印失败率FAILURE RATE 20–30% → < 1%
02

大幅减少试错次数 Fewer trial iterations

在仿真中完成参数迭代,找到最优方案后再实际打印 — 从平均 3–7 次试错降至 1 次。每一次试错都是上万元的粉末与几十个小时。

Iterate in silico, find the optimal window, then print — cutting 3–7 trial prints down to one. Every avoided trial saves tens of thousands in powder and dozens of hours.

试印次数TRIALS PER WIN 3–7 → 1
03

缩短工艺开发周期 Shorter process-dev cycle

云熔智算的 AI 助手 SimuAgent 自动设置参数、自动生成报告。原本需要数周的工艺摸索周期,在 AI 辅助下压缩到数小时。

The SimuAgent assistant auto-configures parameters and writes the report. What used to be weeks of process-window hunting collapses to hours with AI assistance.

开发周期DEV CYCLE 6 wk → ~1 day
04

仿真预测 ↔ 实验 高度一致 Sim predictions match experiment

仿真不是 demo — 每一个预测都拿到实验台上逐项比对。钨倒铜复合材料 EBAM 案例中,仿真与实验达到 90% 一致性。

Simulation isn’t a demo — every prediction is matched against experiment, line by line. In the W/Cu EBAM case, sim and experiment agreed at 90%.

SIM ↔ EXPSIM ↔ EXP 90%
REASON 01 · 世界领先精度World-leading accuracy

屡获大奖
的仿真精度
Award-winning
accuracy

多物理场耦合仿真引擎,覆盖流场、热场、电磁场,从微观晶粒生长到宏观机械性能逐层重现 3D 打印过程。每一个仿真结果都经过实验验证。 A multi-physics coupled simulation engine spanning flow, thermal and electromagnetic fields — reproducing the entire AM process from grain-scale microstructure up to bulk mechanical properties, every result experimentally validated.

  • 仿真精度 82.3%,最大偏差 22.6%Accuracy 82.3% · max deviation 22.6%
  • 2022 / 2025 NIST 增材制造挑战赛 8 个第一8× 1st place · NIST AM Bench 2022 / 2025
  • 200+ 篇同行评议论文支撑Backed by 200+ peer-reviewed publications
了解仿真引擎Explore the engine
实验 vs 仿真 · 应变云图EXP vs SIM · STRAIN OVERLAY R² = 0.978
Experiment vs simulation strain overlay
实验Experiment 云熔智算YunRong
2022 NIST AM Bench · 5× 第一 + 4× 第二2022 NIST AM Bench · 5× 1st + 4× 2nd 2022
2022 NIST AM Bench award plaques
REASON 02 · 全流程覆盖End-to-end coverage

覆盖 3D 打印
全流程
The only platform
covering the full AM pipeline

40+ 仿真功能,10+ 独有。从粉末铺设到熔池演化、微观结构生长、残余应力,到最终零件机械性能 — 一个引擎,一套数据。 40+ simulation functions, 10+ proprietary. From powder bed to melt-pool dynamics, microstructure evolution and residual stress, all the way to final part properties — one engine, one data flow.

  • 支持 SLM / EBM / 粘结剂喷射等主流工艺SLM · EBM · Binder Jetting and more
  • 金属、合金、复合材料、高熵合金Metals · alloys · composites · HEAs
查看功能矩阵See feature matrix
金属 3D 打印全流程仿真 Metal AM end-to-end simulation
REASON 03 · GPU 并行加速GPU acceleration

从一个月
压缩到三天
From a month
down to three days

GPU + CPU 混合并行运算,相同精度下速度提升 10 倍。AI 增强预测把单次推理压缩到毫秒级 — 让仿真不再是瓶颈。 Hybrid GPU + CPU parallelism delivers up to 10× speedup at the same accuracy. AI-enhanced prediction collapses inference to milliseconds — simulation stops being the bottleneck.

  • GPU 集群扩展仿真尺寸上限GPU clusters scale up domain size
  • AI 预测与高保真仿真高度吻合AI predictions match high-fidelity sim
  • 支持云端弹性算力Elastic cloud compute
查看性能基准See benchmarks
同一仿真 · 不同后端SAME SIMULATION · DIFFERENT BACKEND
CPU · 单线程CPU · single thread
~30 天~30 DAYS
RTX 3080 · GPURTX 3080 · GPU
3 天 · 10 倍提升3 days · 10× faster
AI 代理模型AI surrogate
10 毫秒 · 2.6 亿倍提升10 ms · 260M× faster
同等精度下 · 实测 NIST AM Bench 2022 Equal accuracy · NIST AM Bench 2022
REASON 04 · SIMUAGENT AI

对话即仿真
零门槛上手
Conversation in,
simulation out

SimuAgent AI 助手把仿真所需的几十个参数压成一句话需求。自动建模、自动求解、自动出分析报告 — 工程师不必再是 CAE 专家。 SimuAgent collapses dozens of simulation parameters into a single natural-language request. It auto-builds the model, runs the solver, and writes the analysis report — your engineers no longer need to be CAE specialists.

  • 自动参数化、自动网格、自动后处理Auto-parameterize · auto-mesh · auto-postprocess
  • 输出可解释的工艺优化建议Explainable process recommendations
  • 支持企业知识库 + 工艺包嵌入Bring-your-own knowledge base
试用 SimuAgentTry SimuAgent
SIMUAGENT · DEMO ● LIVE
已与全球数十家科研机构、商业公司建立合作 —— Trusted by dozens of research institutes & companies worldwide ——
NUS·SG
清华·THU
中国钢研·CISRI
NIST·USA
A*STAR·SG
上汽·SAIC
航天科技·CASC
NWU·USA
告诉我们你的打印难题Tell us your AM problem

把工艺难题
交给云熔智算。
Hand off your toughest
process problem.

提交一个简短的需求,我们的应用工程师会在 48 小时内回复你 — 评估可行性、给出仿真路径、安排 Demo。 Send a short brief. Our application engineers reply within 48 hours — feasibility check, simulation plan, Demo slot.

  • 01
    不需要任何 CAE 经验 No CAE expertise required 简单描述零件、材料、设备即可 Just describe part · material · machine
  • 02
    NDA 优先 NDA-first 所有 CAD 与工艺数据严格保密 All CAD & process data kept confidential
  • 03
    试点零部件免费 First pilot part free 新客户首件仿真零费用 Zero-cost first simulation for new accounts
需求提交 · DEMAND BRIEF DEMAND BRIEF · 需求提交
✓ 已收到,48 小时内回复。 ✓ Received — we'll reply within 48h.