MA6502 — 增材制造文件格式 + DfAM + 基准测试

知识点类别：3D 打印数据格式 & 面向增材制造设计 & 性能基准测试

包括：STL 格式 / Tessellation（镶嵌）/ STL 缺陷与修复 / AMF / 3MF / DfAM / 仿真建模 / NIST 测试样件

一、中英对照总结

1. STL 格式（STereoLithography / Standard Triangle Language）

• 由 3D Systems 提出，至今仍是 AM 事实标准；
• 文件内容：无序的三角面片列表，每个面片 = 3 个顶点 (X,Y,Z) + 1 条单位法向量（指向“外”）；
• 两种形式：
  • Binary：体积小；
  • ASCII：可读，但文件大；
• CAD 软件（SolidWorks、Creo / Pro-E、AutoCAD 等）可直接 Export 为 STL。

Developed by 3D Systems, STL remains the de-facto standard in AM. It stores an unordered list of triangular facets, where each facet has three vertices (X,Y,Z) and one outward unit normal vector. It has two versions: Binary, which is compact, and ASCII, which is readable but larger. CAD software such as SolidWorks, Creo / Pro-E, and AutoCAD can export directly to STL.

ASCII 片段示例：

facet normal 0.0 1.0 0.0
  outer loop
    vertex 0 0 2
    vertex 0 0 0
    vertex 1 0 2
  endloop
endfacet

2. Tessellation（镶嵌 / 多边形化）

• 定义：用几何形（三角形或四边形）无缝无重叠地逼近原始曲面；
• 常用三角形（3 顶点）或四边形（4 顶点）；
• 分辨率：
  • 高分辨率 → 三角多、表面光滑；
  • 低分辨率 → 三角少、表面粗糙；
• 打印上限：打印件永远不会比 STL 模型更光滑。

Tessellation approximates an original surface with non-overlapping polygons, usually triangles or quadrilaterals. Higher resolution creates more triangles and a smoother surface, while lower resolution uses fewer triangles and gives a rougher surface. A printed part can never be smoother than its STL model.

2.1 Chord Height（弦高）

• 弦高 = 原始曲面与近似多边形网格之间的最大距离；
• 弦高越小 → 近似越光滑；
• 弦高比 = 1 − 弦高 / 弦长；越接近 1 越光滑。

Chord height is the maximum distance between the original surface and the approximate polygon mesh. Smaller chord height means smoother approximation. The chord-height ratio is 1 - chord height / chord length; the closer it is to 1, the smoother the approximation.

3. STL 的优缺点

优点

• 表示 3D CAD 数据的简单方法；
• 事实标准、绝大多数 CAD / AM 支持；
• 简单形状文件小、准确。

Pros: STL is a simple way to represent 3D CAD data. It is a de-facto standard supported by most CAD and AM systems. For simple shapes, STL files can be small and accurate enough.

缺点

• 冗余：重复顶点 / 边，文件比原始 CAD 大许多倍；
• 不同 CAD 导出的 STL 经常有缺陷，需要修复软件（如 Magics），拖慢生产周期；
• 大 STL 切片可能需数小时；
• 仅能存几何 — 丢失颜色、纹理、材料；
• 扁三角形难以扩展到复杂几何（曲面、微结构）；
• 复杂曲面相交、数值误差难控；
• 典型复杂对象需数百万三角（10 cm 球 @ 10 µm 精度 → 约 20,000 三角）。

Cons: STL is redundant because vertices and edges are repeated, so files can be much larger than the original CAD model. STL files exported from different CAD systems often contain defects and require repair software such as Magics, slowing production. Large STL files may take hours to slice. STL stores geometry only, so colour, texture, and material information are lost. Flat triangular facets are inefficient for curved surfaces and micro-geometry, and complex intersections or numerical errors are hard to control. Complex objects may require very large numbers of triangles.

简短讨论：STL 的核心矛盾是“简单通用”与“信息贫乏”。它只用三角面片描述外形，所以软件容易读写、设备兼容性强，简单零件也能保持较小文件；但同一个顶点会在多个三角形中重复保存，曲面越复杂就越需要大量三角形，文件会膨胀，切片也会变慢。更重要的是，STL 不保存颜色、纹理、材料、梯度或晶格信息，所以它适合基础几何传递，不适合现代多材料、多色和微结构 AM。考试讨论时可写：STL 仍流行不是因为最好，而是因为生态成熟、兼容性高；新格式 AMF/3MF 的出现正是为了解决这些限制。

Brief discussion: The key trade-off of STL is “simple and universal” versus “information-poor.” Because STL describes only outer geometry with triangular facets, it is easy for software and machines to read, and simple parts can stay compact. However, vertices are repeatedly stored across facets, so curved or complex parts require many triangles, leading to large files and slower slicing. More importantly, STL cannot store colour, texture, material, gradients, or lattice information, so it is suitable for basic geometry exchange but weak for modern multi-material, full-colour, and microstructured AM.

4. STL 常见缺陷（四大类）

4.1 缺面 / 缝隙（Missing Facets / Gaps）

• 曲率大的面相交处 Tessellation 易在边缘留空；
• 切片时会有“游离扫描矢量（stray scan vectors）”、激光无边界停 → 打印失败；
• 修复：用 Magics 等软件自动补面。

Missing facets or gaps often appear at intersections of highly curved tessellated surfaces. During slicing, they may create stray scan vectors because the laser has no clear stopping boundary, causing print failure. They are commonly repaired by automatic facet filling in software such as Magics.

4.2 退化三角面（Degenerate Facets）

• 三个顶点虽然不同，但三边共线 → 三角形退化成线；
• 法向量无效，但保留了曲面交接处的拓扑信息；
• 通常源于 stitching。

Degenerate facets have three different vertices, but the vertices are collinear, so the triangle degenerates into a line. The normal vector is invalid, although the facet may still preserve topological information at surface junctions. This usually comes from stitching.

4.3 重叠三角（Overlapping Facets）

• 顶点用浮点表示，数值四舍五入时在容差设置过松时产生；
• 同一位置多个三角叠加。

Overlapping facets are caused by floating-point vertex coordinates and loose tolerance settings during numerical rounding. The result is multiple triangles occupying the same position.

4.4 非流形（Non-Manifold）

• 一个有效模型：每条边只连接 2 个面；
• 三种非流形：
  • 非流形边：4 个面共享一边；两立方体共享一棱；
  • 非流形点：两实体仅靠一顶点连接；
  • 非流形面：面两侧无封闭体；
• 视觉上“看似封闭”，切片时却“开放” → 3D 打印无法处理（可做可视化但不能打印）。

For a valid model, each edge should connect exactly two facets. Non-manifold geometry includes non-manifold edges, where more than two facets share one edge; non-manifold points, where two solids touch only at one vertex; and non-manifold faces, where the face does not bound a closed volume. Such models may look closed visually, but the slicer treats them as open, so they cannot be handled reliably for 3D printing.

5. 有效 vs 无效镶嵌模型

• 有效：无缺陷，可直接切片 → 生成 2D 层 → 1D 扫描线（激光开关信号）；
• 无效：必须先修复；在无预定停止边界时，扫描线飞出模型边界 → 错误打印。

A valid tessellated model has no critical defects and can be sliced into 2D layers and then converted into 1D scan lines or laser on/off commands. An invalid model must be repaired first; otherwise, scan lines may leave the model boundary when no stopping boundary is defined, leading to incorrect printing.

6. STL 修复

• 通用 + 专用解决方案；
• 检查所有三角面有效性；
• 破洞修复耗时且繁琐；
• 最大难点：修复必须匹配设计者意图 → 多数情况是“受过训练的猜测”；
• 市场上很多商业修复软件（Magics、Netfabb 等）。

STL repair uses both generic and specialised solutions. It checks the validity of all triangular facets. Hole repair is often time-consuming and tedious. The hardest part is that the repair must match the designer’s intent, which is often an educated guess. Commercial tools include Magics and Netfabb.

7. 为什么需要新格式？

• 1980 年代：单材料、均匀形状；
• 现在：多材料、多色、梯度、晶格 / 多孔、纹理、微结构、更高分辨率；
• STL 完全无法描述这些新特性 → 催生 AMF 和 3MF。

In the 1980s, AM mainly involved single-material and homogeneous shapes. Today, AM needs multi-material, full-colour, graded, lattice/porous, textured, microstructured, and higher-resolution descriptions. STL cannot describe these new features, so AMF and 3MF were created.

8. AMF（Additive Manufacturing Format）

• ISO/ASTM 52915 标准（2020 年 v1.2）；
• 基于 XML，存为纯文本或 ZIP 压缩；
• 支持能力：
  • 曲面三角形（curved triangles）：法向量起点不必在三角形平面内 → 更少三角即可光滑 → 解决大 STL 文件问题；
  • 颜色：RGB + 透明度，可嵌套赋给顶点 / 三角 / 体积 / 对象 / 材料；
  • 纹理：几何缩放 + 像素强度 + 上色（类似图形学纹理映射，不是物理凹凸）；
  • 材料：不同体积可指定不同材料，设计阶段即可确定；
  • Material variants（算子）：可应用晶格 / 多孔 / 梯度 / 周期性 / 随机算子 → 医疗 / 航空 AM 的理想选择；
  • 多物体、2D & 3D 纹理、元数据；
• 未来将支持公差、表面粗糙度、涂层、支撑结构、数字签名与内容保护；
• 支持旧 STL（需转换）；
• 已经成熟并在市场上运行。

AMF is the ISO/ASTM 52915 standard, with v1.2 in 2020. It is XML-based and can be stored as plain text or ZIP. It supports curved triangles, colours, textures, material assignment, material variants, multiple objects, 2D/3D textures, and metadata. Curved triangles allow smoother geometry with fewer triangles, helping solve the large-STL-file problem. Material variants can describe lattice, porous, graded, periodic, and stochastic structures, which is useful for medical and aerospace AM. Future extensions include tolerances, surface roughness, coatings, support structures, digital signatures, and content protection. It supports legacy STL through conversion and is already mature in the market.

9. 3MF（3D Manufacturing Format）

• 由 3MF Consortium（联合开发基金会项目）发起；
• 设计目标：
  • Complete：一个存档容纳模型 + 材料 + 属性；
  • Human readable：OPC / ZIP / XML 通用结构；
  • Simple：规范短小清晰；
  • Extensible：XML 命名空间允许公有 / 私有扩展；
  • Unambiguous：一致性测试确保从数字到物理结果一致；
  • Free：永久免版权 / 专利 / 许可。

3MF was initiated by the 3MF Consortium. Its design goals are: Complete, meaning one archive contains model, material, and properties; Human readable, using OPC / ZIP / XML; Simple, with a short and clear specification; Extensible, using XML namespaces for public and private extensions; Unambiguous, with conformance tests to ensure digital-to-physical consistency; and Free, meaning royalty-free and licence-free.

10. AMF vs 3MF

AMF is stronger in curved triangles and graded/lattice scripting. 3MF is stronger in compact ZIP storage, OS integration, STL compatibility, and digital signatures. Both are XML-based; AMF is mature, while 3MF is developing quickly.

11. DfAM（Design for Additive Manufacturing，面向增材制造的设计）

将 DfM/DFMA 应用到 AM 的设计学。

DfAM applies DfM/DFMA principles to additive manufacturing design.

11.1 与传统 DfM 的区别

• 传统 DfM 原则：易装配、标准件、减少零件数、减少加工工序；
• 3D 打印制约少 → 允许更以优化为导向的设计流程（复杂几何几乎免费）。

Traditional DfM emphasizes easy assembly, standard parts, fewer components, and fewer machining steps. Because 3D printing has fewer manufacturing constraints, it allows a more optimization-driven design workflow, where complex geometry is nearly free compared with traditional manufacturing.

11.2 DfAM 三大研究方向

1. Component 设计（决策支持）— 哪些零件用 AM 制造在经济 / 生态上有收益？
2. Part 设计：

• 拓扑优化（Topology Optimization）— 移除不受力区域，留最优力流路径；
• 晶格结构设计（Lattice）— nTopology、Autodesk Within 等；
• 设备 / 工艺专用设计规则（角度、最小特征、孔径、悬垂等，Hubs 等有在线手册）。

3. Process 设计：

• 支撑结构优化与位置；
• 构建方向（可自支撑化）；
• 机器参数 / 软件设置；
• 成本计算；
• 仿真（分析、数值、经验）。

The three DfAM research directions are: Component design, deciding which parts benefit economically or ecologically from AM; Part design, including topology optimization, lattice design, and machine/process-specific rules; and Process design, including support optimization, build orientation, machine parameters, software settings, cost calculation, and simulation.

11.3 Part Decomposition（零件分解）

• 与 Part Consolidation（合并）相反 — 把一个件分拆成多个子装配；
• 三个动机：
  1. 适配构建空间（设备尺寸限制）；
  2. 减少构建时间和成本；
  3. 提升质量（某些方向打印更好）。
• 研究方向：表面质量、装配界面、分解准则、Kinematics-aware 朝向、双工艺分解等。

Part Decomposition is the opposite of Part Consolidation: it splits one part into multiple sub-assemblies. The motivations are to fit the build envelope, reduce build time and cost, and improve quality by choosing better print orientations. Research topics include surface quality, assembly interfaces, decomposition criteria, kinematics-aware orientation, and dual-process decomposition.

12. 仿真与建模（Simulation & Modelling）

• FEA（有限元） 影响 FDM 填充设计 → 按受力生成填充；
• CFD（计算流体） 模拟 LPBF 熔池动力学（例 NTU 研究）；
• Process–Structure–Property 关系研究 → 数据驱动方法；
• 目标：把“试错”变为“预测”。

Examples include FEA-driven FDM infill, where infill is generated according to load paths; CFD simulation of LPBF melt-pool dynamics, such as NTU research; and Process–Structure–Property relationship modelling using data-driven methods. The goal is to turn trial-and-error into prediction.

13. DfAM 相关标准

• ISO/ASTM 52910:2017(E) — DfAM 一般指导；
• ISO/ASTM 52911-1:2019(E) — 金属激光 PBF 设计；
• ISO/ASTM 52911-1:2019(E) — 聚合物激光 PBF 设计；
• ASTM F3413-2019 — 金属 DED 设计。

Key DfAM standards include ISO/ASTM 52910:2017(E) for general DfAM guidance, ISO/ASTM 52911-1:2019(E) for metal laser PBF design, ISO/ASTM 52911-1:2019(E) for polymer laser PBF design, and ASTM F3413-2019 for metal DED design.

14. Benchmarking（性能基准测试）

14.1 两类方法

1. 直接测量机器 / 工艺特性（轴、激光等）—— 难：难以触及轴、传感器会干扰过程；
2. 测量制造出的测试样件（test piece）—— AM 中更重要。

There are two methods: directly measuring machine or process characteristics, such as axes and lasers, which is difficult because axes are hard to access and sensors may disturb the process; or measuring the manufactured test piece, which is more important in AM.

14.2 标准测试样件（Standard Test Artifact）

• 量化评估设备 / 工艺性能；
• 不同设备打同一测试件可横向对比；
• 若设计得当，可测极限、发现改进点，并用于机器补偿参数。

A standard test artifact quantitatively evaluates machine or process performance. Printing the same artifact on different machines enables cross-machine comparison. If designed well, it can measure limits, identify improvement areas, and support machine compensation parameters.

14.3 NIST 测试样件

包括以下典型特征：

• 菱形基座；
• 阶梯和阶梯立面；
• 销钉和孔；
• 细小特征；
• 中心孔 + 圆柱；
• 斜坡；
• 侧向特征；
• 顶面、外缘等。

The NIST test artifact includes typical features such as a diamond base, staircases and staircase walls, pins and holes, fine features, a central hole and cylinder, ramps, lateral features, top surfaces, and outer edges.

14.4 其他 Benchmarking 议题

• 不同 AM 工艺比较（SLM / EBM / SLS / MJF 对比）；
• 监管挑战：FDA（医疗）、FAA / EASA（航空）；
• Gartner、NIST 路线图跟踪。

Other benchmarking topics include comparison across AM processes such as SLM / EBM / SLS / MJF, regulatory challenges from FDA for medical applications and FAA / EASA for aerospace applications, and roadmap tracking from Gartner and NIST.

二、速记要点 / Quick Recall

• STL = 三角面片 + 单位法向量；两种格式：Binary / ASCII。
• 弦高 Chord Height 越小 → 近似越光滑；打印永不比 STL 平滑。
• STL 四大缺陷：Missing facets、Degenerate、Overlapping、Non-manifold；Non-manifold 是一条边只能连两面。
• AMF（ISO/ASTM 52915）强项：曲面三角形 + 梯度 / 晶格 / 多孔脚本。
• 3MF 强项：紧凑 ZIP、数字签名、原生兼容 STL、Windows 管线。
• DfAM 三层：Component（要不要 AM）→ Part（拓扑 / 晶格 / 规则）→ Process（支撑 / 朝向 / 仿真）。
• Part Consolidation（合并）vs Part Decomposition（分解）。
• Benchmarking：NIST 标准测试样件 是跨设备对比的关键工具。
• STL = triangular facets + unit normal vector; two formats: Binary / ASCII.
• Smaller Chord Height means smoother approximation; a printed part can never be smoother than its STL model.
• Four STL defects: Missing facets, Degenerate facets, Overlapping facets, Non-manifold; for manifold geometry, each edge should connect exactly two facets.
• AMF (ISO/ASTM 52915) strengths: curved triangles + graded/lattice/porous scripts.
• 3MF strengths: compact ZIP, digital signature, native STL compatibility, Windows print pipeline.
• Three DfAM levels: Component (whether to use AM) → Part (topology/lattice/rules) → Process (support/orientation/simulation).
• Part Consolidation versus Part Decomposition.
• Benchmarking: the NIST standard test artifact is key for cross-machine comparison.