上个月有个做冷板框架的客户，拿着一块改性尼龙注塑的支架来找我们。
框架是用来固定服务器冷板的，边缘还带着定位销孔。他把它平放在桌上，手指一推："你们看，装上怎么一边翘起来了？是料太软了吧。"
我们没先答"软不软"，先把件翻过来看了浇口位置，又问了玻纤含量。答案回来——PA66-GF30，浇口开在长边一头。
这件事把冷板结构件选料的核心说透了：翘曲不是料软了，是玻纤排错了队。 这一件要的从来不是"更结实"，是"刚得住还不翘"。
一、冷板框架这一件，到底在干什么
先说清它和冷板本体的关系。
冷板本体是导热的，必须金属（铜或铝），尼龙不导热，这一块从来不是尼龙的活。尼龙做的是冷板周围的框架、支架、定位件、分流器壳体、固定夹——它们不导热，但要干三件事：
撑住冷板，长期预紧不变形；定位管路和快接头，尺寸要稳；抗住装配和热循环的应力，不开裂。
这三件事里，没有一件是"强度够不够"能单独回答的。它要的是刚性、尺寸稳定、低翘曲三样同时拿到——而这恰恰是相互牵制的。
一句话：冷板结构件选料，考的不是拉伸强度，是"装上去平不平、十年后还稳不稳"。
把"平不平"换成数字：带定位销孔的框架，平面度常按 0.5 mm 量级控制；而 PA66 吸湿后一个 200 mm 边能漂 1 mm 量级——也就是说，干态平平整整的件，装机房高湿半年，可能直接超出装配公差。这就是为什么尺寸稳定要单列，不能靠"强度够"兜底。
二、工况六维：这件被什么约束
按六个维度摊开。
温度。 冷板式回路 40–55℃，框架贴着冷板，局部可能到 70–80℃。不算高，但热循环频繁——开机升温、关机降温，一天几轮。
载荷。 主要是装配预紧力和自身重力，载荷不高，但长期。振动来自机房风力和搬运。
介质。 框架不直接泡冷却液，但靠近接头和管路，有溅湿和凝露可能。所以阻燃和低烟仍是硬约束。
寿命。 十年起步。判据不是"断没断"，是"第十年平面度和定位尺寸还在不在公差带里"。
十年换成小时，约八万七千小时连续运行。框架的判据不是"能不能撑到十年"，是"第八万小时平面度还在不在公差里"。很多件强度十年后仍有余量，却因慢漂和蠕变，第七年定位就偏了——所以看寿命要看尺寸寿命，不是强度寿命。
外观与洁净。 机房低烟低毒，外观件看色差和长期黄变。
合规。 阻燃 UL94 V0，不少场合无卤；靠近电器看 GWIT。
六维里温度、载荷、寿命都给了具体数字，这就是件级精度。
温度那行再细化：框架贴着冷板，局部温度可能到 70–80℃，已经逼近普通 PA66 舒适区边缘。热循环一天几轮，每次升降十几度，界面应力反复拉扯——十年下来，累计循环次数以万计，这对材料是慢刀子。
三、三条玻纤路线，刚性和翘曲要一起算
把候选路线并排放，看"刚性"和"翘曲"两列怎么走。
| 路线 | 组成 | 刚性 | 翘曲 / 尺寸 | 代价 |
|---|
| PA66-GF30/40 | 脂肪族玻纤增强 | 好，弯曲模量 8–11 GPa | 吸湿翘曲、各向异性 | 成本低、工艺成熟 |
| PA6T/PA9T-GF | 半芳香族玻纤增强 | 好，且耐温高 | 低吸湿、尺寸最稳 | 贵、加工窗口窄 |
| PA66-GF + 矿物混杂 | 玻纤与矿物/玻微珠混填 | 略降 | 各向同性、翘曲最小 | 刚性换来的平衡 |
三条路线没有"谁更好"，只有"哪条线更紧"。
PA66-GF30 是主流——刚性够、便宜、好打。但它的吸湿率在 8% 上下，泡水或高湿环境里尺寸会漂，各向异性收缩还会翘。 对平面度要求松的支架能容忍，对带定位销孔的框架就悬。
为什么吸湿高的料翘得凶？尼龙吸水后分子链被撑开，整体膨胀；玻纤不吸水，于是沿纤维方向和垂直方向胀得不一样，板就卷。这不是"软"，是"各向胀得不齐"。理解到这一层，治翘曲才会去动浇口和混杂，而不是一味加玻纤。
PA6T-GF 吸湿低、尺寸稳、耐温高一档，适合精密定位件。代价是单价和加工难度。
用量级比：PA6T 吸湿通常只有 PA66 的几分之一，同样 200 mm 定位边，湿态漂移可能从 1 mm 量级降到 0.2–0.3 mm 量级。对松公差支架这差别无所谓，对带销孔框架就是"装得进"和"装不进"的差别。
矿物混杂那一行是治翘曲的——牺牲一点刚性，换来各向同性，大平板件特别受用。它不"更结实"，但"更平"。
一句话：冷板结构件，刚性看玻纤，翘曲看混杂和浇口，尺寸稳定看基材吸湿。三件事分开抓，才抓得准。
四、选型判据表：平面度和尺寸稳定是主角
把约束落成可核对指标。下表门限是方向性建议，不是验收标准——实际由项目、工况和实测定。
| 指标 | 方向性门限 | 验证方法 / 标准 | 常见失效 | 通行解法 | 对应助剂体系 |
|---|
| 弯曲模量（刚性） | 参考 8–11 GPa 量级 | ISO 178 | 受力变形、松 | 玻纤增强 | 偶联剂（界面增强） |
| 翘曲 / 平面度 | 大板平面度按件定，常 ≤ 0.5 mm | 三坐标 / 平台法 | 装配干涉、翘边 | 低翘曲体系 + 浇口优化 | 成核剂（控制结晶） |
| 吸湿尺寸稳定 | 湿态与干态差可控 | ISO 62 + 量测 | 定位偏、卡不进 | 低吸湿基材 | 材料本征，不靠助剂 |
| 热膨胀匹配 | 与铝冷板 CTE 差可控 | 热循环 + 尺寸 | 嵌件处开裂 | 选近 CTE 基材 / 结构 | 偶联剂（界面） |
| 长期耐温 | 80℃×1000 h 强度保持 | ISO 527 | 发白、脆化 | 热稳定体系 | 抗氧剂（热氧） |
| 嵌件结合 | 热循环后无裂、拉拔够 | 温度循环 + 拉拔 | 嵌件松、漏 | 咬合结构 + 基材选配 | 偶联剂（界面） |
| 阻燃 | UL94 V0，优先无卤 | UL94 / GWIT | 安规返工 | 无卤阻燃体系 | — |
怎么用这张表：刚性一行和翘曲一行要分开打。只盯模量不盯平面度，框架装上去就翘，强度再高也没用。 带定位孔的件，尺寸稳定那行权重最高。
一个实用建议：带定位孔的件，把"湿态平面度"和"湿态定位尺寸"两条单独列红线，不合并进笼统的"尺寸稳定"。合并了，吸湿漂移容易被平均掉，装机后才暴露。
一个提醒：翘曲这一项，浇口位置和流道平衡的影响往往不亚于材料本身。大平板件先动浇口，常常比换料更省事——玻纤是按流动方向排队的，队列乱了，件就翘。
这也解释了为什么我们不让客户一上来就换高玻纤。浇口一改、模温一提，同一支料平面度可能直接回来；先动料再动模，等于把工艺问题甩给材料背。
五、四条常见误判，和真实根因
判反一：翘曲就是料软，换高玻纤。
这是最可惜的一类。客户一看框架翘，第一反应是"加玻纤、换高刚性料"。翘的根源常在各向异性收缩——玻纤沿流动方向排，垂直方向缩得少，长板件一头缩多一头缩少，就翘了。 先动浇口、调模温、做低翘曲混杂，往往比加玻纤有效。我们见过同料同模，浇口一改，平面度直接回来。
判反二：用干态尺寸做定位设计。
尼龙吸湿会涨。PA66 吸湿 8%，一个 200 mm 长的定位边能变 1 mm 量级。用干态尺寸定的销孔，装机房高湿环境跑半年，定位就偏了。 尺寸报告必须按调湿态出，定位公差按湿态留。
判反三：忽略嵌件和冷板的热膨胀差。
铝的线膨胀系数约 23×10⁻⁶/K，PA66 约 70–80×10⁻⁶/K，差三倍多。温度循环之下，界面应力越积越大，嵌件周围先裂。 这不是材料强度不够，是热膨胀没配。处理靠咬合结构和选近 CTE 的基材，早期就要定。
判反四：阻燃剂加完就完事。
无卤阻燃体系加量大，某些阻燃剂或抗氧剂在长期热循环下会往表面迁移，出现析出、发雾、黄变。这是配方侧的坑——加工温度超了耐温，或助剂耐温不够，表面就脏。 看到黄变析出，先查助剂耐温，别急着换基材。
一条时间线（行业常见的嵌件开裂路径）：框架注塑、常温装配合格 → 机房运行热循环 → 嵌件周围应力累积 → 半年到一年出现微裂 → 某次搬运振动扩大 → 追溯是 CTE 差加预紧力。问题在选材阶段就埋了，只是慢。
拿到冷板框架询盘，我们通常会追问三句：第一，定位精度按干态还是湿态出图；第二，和什么金属冷板配、热膨胀差有没有算；第三，年用量够不够摊开模和验证。三句答不清，料定不稳。
六、加工与验证：翘曲和嵌件要分开盯
冷板结构件是注塑件，两个坑分开控。
干燥。 尼龙必烘，PA66 含水超 0.15% 注塑即可能水解降解，强度掉、内应力大。干燥按实测含水率定窗口，南方梅雨季尤其要除湿干燥机。
浇口与流道。 大平板件浇口位置决定流动平衡，直接决定翘曲。先动浇口，再谈换料。
模温。 模温低，表面浮纤、内应力大、翘曲加剧。玻纤件模温通常要提到 110–120℃ 量级。
这 110–120℃ 看着高，却是玻纤件的常规区间；模温低 20℃，浮纤和翘曲可能明显恶化。模温不是"越高越好"，是"够让玻纤被树脂包住、内应力小"。
嵌件。 金属嵌件预热的温差、咬合结构、注塑包覆顺序，都影响界面。嵌件周围做局部加强筋，把应力摊开。
验证顺序。 建议这样排：
1. 材料级：弯曲模量、热氧保持率
2. 尺寸级：干态与湿态平面度、定位尺寸
3. 嵌件级：温度循环 + 拉拔复测
4. 阻燃级：UL94 / GWIT
5. 系统级：上整机前最后验
顺序不能换。 前一项没过就往下走，后面数据没有解释意义。
有个常被跳过的细节：嵌件预热。金属嵌件不预热直接打，和尼龙温差大，界面一冷就缩、易脱粘；预热到接近模温，咬合才牢。这道工序省了，前面尺寸全过，热循环一来嵌件先松。
七、边界：什么时候不该用改性尼龙
这一段可能比前面更值钱。
其一，冷板本体要导热。 那是金属（铜/铝）的活，尼龙不导热，别往这上面想。
其二，长期温度超过 150℃ 的靠近热源位置。 普通 PA66 体系不够，要半芳香族或 PPS 类。
其三，超大载荷的主承力框架。 尼龙的模量和蠕变摆在那里，这类位置金属更稳。
其四，年用量小到摊不平注塑模和验证。 带定位孔的精密件要开模、做浇口优化、跑热循环，年用量几百个不成立。
其五，要求零析出可见、贴近电器。 这类要把阻燃析出预期提前对齐，必要时换更洁净体系。
把这五条写在前头，不是劝退，是省时间。样品阶段平平整整、量产因翘曲或嵌件裂回退的项目，见过不止一个——回退代价比当初不做高得多。
还有本经济账：带定位孔的精密框架，开模加浇口优化加热循环验证，一次性投入往往按万元计。年用量几百个的话，摊到每件比料钱还高。立项前先算清用量，比先选牌号更实在。
八、换料风险清单（从原方案换到玻纤尼龙冷板结构件，要动什么）
| 环节 | 要动什么 | 容易漏的点 |
|---|
| 模具 | 浇口按流动平衡重定，不套旧位 | 玻纤取向导致翘曲 |
| 干燥 | 按实测含水率定窗口 | 回用料带入水分 |
| 料温 / 模温 | 玻纤件模温提到 110–120℃ 量级 | 只按牌号推荐值给 |
| 保压与脱模 | 嵌件周围加强、顺序定 | 嵌件预热温差 |
| 调湿 | 尺寸按湿态出，不按干态 | 干态平、湿态偏 |
| 色差 | 外观件色板提前确认 | 阻燃体系黄变预期 |
| 验证顺序 | 材料 → 尺寸 → 嵌件 → 阻燃 → 系统 | 前一项未过就往下走 |
九、一页纸汇报表（给要向上汇报的人）
`
项目：冷板框架 / 液冷板结构件 · 材料路线评估
结论方向：玻纤尼龙可作为候选，能否落地取决于三项前置
一、必须守住的三条
1. 平面度按湿态定，不按干态
2. 嵌件热膨胀差早期配，不靠材料硬扛
3. 翘曲先动浇口，再谈换料
二、前置条件（任一不满足则建议暂缓）
· 长期工作温度 ≤ 80℃ 量级（局部另核）
· 与金属冷板有热膨胀匹配方案
· 年用量足以摊薄注塑与验证投入
三、下一步动作
1. 做干湿态平面度与定位尺寸
2. 做嵌件温度循环 + 拉拔
3. 定浇口与流道平衡方案
风险提示：本路线主要不确定性在翘曲与吸湿尺寸，不在初始强度。
`
十、读者常问的两句
问：和进口玻纤尼龙差在哪？
只讲两件能对照的事：同一指标，看它标没标测试状态（干态还是湿态）；同一件上，看它给没给翘曲和尺寸稳定的长期数据。结构件对状态极敏感，状态不明的数字不宜直接比。有些支架走国产 PA66-GF 路线已经成熟，有些精密定位件仍建议半芳香族——具体到你的平面度要求，要看温度和湿度两样。
问：PA66-GF30 够不够，要不要上半芳香族？
看位置和精度。纯支撑、平面度松的支架，PA66-GF30 够用也划算；带定位销孔、贴冷板热循环、要求十年尺寸稳的，PA6T-GF 的低吸湿更值当。 升级前先做三件事：确认湿态尺寸真超差、确认爬电或结构余量已优化、确认温升在半芳香族舒适区。三件做完再升，钱花得踏实。
做改性尼龙：PA6、PA66、PA46、PA11、PA12、PA6T、PA9T、尼龙合金。
做改性 PPO、PPS、热塑性弹性体。
做各大化工巨头的尼龙树脂、副牌料、大包料。
另：长期收尼龙原料、水口回料与各类尼龙废料，有正规处置渠道。
Last month, a customer who makes cold plate frames came to us with a bracket made of modified nylon injection molding.
The frame is used to secure the server cold plate, with alignment pin holes at the edges. He laid it flat on the table and pushed it with his finger: 'Look, why is one side lifting up when installed? Is the material too soft?'
We didn't first ask 'soft or not'; we flipped the part over to check the gate position and then asked about the fiberglass content. The answer came back—PA66-GF30, with the gate at one end of the long side.
This matter explains the core of selecting materials for cold plate structural components: warping isn't because the material is too soft, it's because the fiberglass was misaligned. What this really requires is never 'stronger,' but 'rigid enough to stay straight without warping.'
1. What exactly is this cold plate frame doing?
First, clarify its relationship with the cold plate itself.
The cold plate itself conducts heat and must be made of metal (copper or aluminum). Nylon does not conduct heat; this part has never been made of nylon. Nylon is used for the frame, bracket, positioning part, manifold housing, and fixing clip around the cold plate—they do not conduct heat, but they have to do three things:
Support the cold plate, maintain long-term preload without deformation; position pipelines and quick connectors, dimensions must be stable; withstand assembly and thermal cycling stress without cracking.
Among these three things, not a single one can be answered solely by 'is the strength sufficient.' What is needed is rigidity, dimensional stability, and low warping all at the same time — and these are precisely mutually constraining.
In a word: when selecting materials for cold-formed structural components, what matters is not tensile strength, but whether 'it sits flat when installed and remains stable ten years later'.
Replace 'flat or not' with numbers: for a frame with locating pin holes, flatness is usually controlled at the 0.5 mm level; whereas after absorbing moisture, a PA66 part with a 200 mm edge can warp by about 1 mm — that is to say, a perfectly flat dry part may exceed assembly tolerances after being in a high-humidity equipment room for half a year. This is why dimensional stability must be specified separately and cannot be covered by 'strength is enough'.
2. Six-dimensional working condition: What constraints this part
Spread out across six dimensions.
Temperature. Cold plate circuit 40–55°C, with the frame touching the cold plate, locally it may reach 70–80°C. Not very high, but the thermal cycling is frequent — heating up when turned on, cooling down when turned off, several cycles a day.
Load. Mainly the assembly pretension and self-weight, the load is not high, but long-term. Vibration comes from the wind in the machine room and handling.
Medium. The frame does not directly soak in the coolant, but it is close to joints and pipelines, so there is a possibility of splashing and condensation. Therefore, flame retardancy and low smoke are still strict constraints.
Lifespan. Starting from ten years. The criterion is not 'whether it breaks or not', but 'whether the flatness and positioning dimensions in the tenth year are still within the tolerance range'.
Ten years converted into hours is about 87,000 hours of continuous operation. The criterion for the framework is not 'whether it can last ten years,' but 'whether the flatness at the 80,000th hour is still within tolerance.' Many components still have strength reserves after ten years, but due to slow drift and creep, the positioning may be off by the seventh year — so when considering lifespan, you have to look at dimensional lifespan, not strength lifespan.
Appearance and cleanliness. The machine room is low smoke and low toxicity, and the appearance parts are checked for color difference and long-term yellowing.
Compliant. Flame retardant UL94 V0, halogen-free in many cases; check GWIT near electrical equipment.
In six dimensions, specific numbers are given for temperature, load, and lifespan, which is at the component-level accuracy.
Further refine the temperature line: the frame is attached to the cold plate, and the local temperature may reach 70–80℃, already close to the edge of the comfort zone of ordinary PA66. The thermal cycle occurs several times a day, with temperature rising and falling by more than ten degrees each time, repeatedly pulling the interface stress—over ten years, the cumulative number of cycles amounts to tens of thousands, which is like a slow knife on the material.
3. For the three fiberglass routes, rigidity and warpage need to be calculated together
Place the candidate routes side by side and see how the 'rigidity' and 'warping' columns go.
| Route | compose; consist of | Rigidity | Warp / Dimensions | Cost |
|---|
| PA66-GF30/40 | Aliphatic glass fiber reinforced | Okay, bending modulus 8–11 GPa | Moisture absorption warping, anisotropy | Low cost, mature technology |
| PA6T/PA9T-GF | Semi-aromatic glass fiber reinforced | Good, and high temperature resistant | Low moisture absorption, most dimensionally stable | Expensive, narrow processing window |
| PA66-GF Mineral mixed | Glass fiber mixed with mineral/glass microspheres | Slightly decrease | Isotropic, minimal warping | Balance achieved through rigidity |
There is no 'which is better' among the three routes, only 'which route is tighter'.
PA66-GF30 is mainstream—it’s rigid enough, cheap, and easy to process. But its moisture absorption is around 8%, so in water or high-humidity environments, dimensions can shift, and anisotropic shrinkage can cause warping. Brackets with loose flatness requirements can tolerate this, but frames with positioning pin holes are problematic.
Why do materials with high moisture absorption warp severely? After nylon absorbs water, its molecular chains expand, causing overall swelling; glass fibers do not absorb water, so the expansion along the fiber direction and perpendicular to it is uneven, and the sheet warps. This is not 'softness,' but 'uneven expansion in different directions.' Understanding this level helps in controlling warping by adjusting the gate and mixing, rather than blindly adding more glass fiber.
PA6T-GF has low moisture absorption, stable dimensions, and high temperature resistance, making it suitable for precision positioning components. The trade-off is its unit price and processing difficulty.
Comparison in magnitude: The moisture absorption of PA6T is usually only a fraction of that of PA66. Similarly, for a 200 mm locating edge, the wet-state drift may decrease from the order of 1 mm to the order of 0.2–0.3 mm. This difference doesn't matter for loose-tolerance brackets, but for frames with pin holes, it is the difference between 'fitting' and 'not fitting'.
The row about mineral mix is for correcting warping — sacrificing a bit of rigidity in exchange for isotropy, which is especially useful for large flat parts. It is not 'stronger,' but 'flatter.'
In a nutshell: For cold plate structural parts, rigidity depends on fiberglass, warping depends on fillers and gate design, and dimensional stability depends on the substrate's moisture absorption. Handle these three aspects separately to manage them accurately.
4. Selection Criteria Table: Flatness and Dimensional Stability are the Main Focus
Turn constraints into verifiable indicators. The thresholds in the table below are directional suggestions, not acceptance standards — the actual values are determined by the project, working conditions, and measurements.
| Indicator | Directional Threshold | Verification Method / Standard | Common Failures | Common solution | Corresponding auxiliary agent system |
|---|
| Bending modulus (rigidity) | Refer to the 8–11 GPa range | ISO 178 | Deformation under stress, loose | Glass fiber reinforced | Coupling agent (interface enhancement) |
| Warp / Flatness | The flatness of large panels is determined per piece, usually ≤ 0.5 mm | Coordinate Measuring Machine / Platform Method | Assembly interference, warping | Low warpage system Gate optimization | Nucleating agent (crystallization control) |
| Moisture absorption dimensional stability | The difference between wet and dry states is controllable | ISO 62 Measurement | Positioning is off, can't fit in | Low moisture-absorbing substrate | Intrinsic material properties, without relying on additives |
| Thermal expansion matching | CTE difference with aluminum cold plate is controllable | Thermal cycle Dimensions | Cracking at the insert | Choose a substrate/structure with a nearby CTE | Coupling agent (interface) |
| Long-term temperature resistance | 80℃ × 1000 h strength retention | ISO 527 | Pale and brittle | Thermally stable system | Antioxidant (Thermal Oxidation) |
| Insert assembly | No cracks and sufficient pull after thermal cycling | Temperature cycling Drawing | Loose or leaking insert | Occlusal structure Substrate selection | Coupling agent (interface) |
| Flame retardant | UL94 V0, halogen-free preferred | UL94 / GWIT | Safety standard rework | Halogen-free flame retardant system | — |
How to use this table: Rigid rows and warped rows should be recorded separately. Focus on modulus, not flatness; once the frame is installed, it will warp, and no matter how strong it is, it won't help. For parts with locating holes, the dimensionally stable row has the highest weight.
A practical suggestion: For parts with locating holes, list 'wet flatness' and 'wet locating dimensions' separately in red, and do not combine them into the general 'dimensional stability.' If combined, moisture-induced warping can easily be averaged out and only become apparent after assembly.
A reminder: Regarding warpage, the influence of gate location and runner balance is often no less than that of the material itself. For large flat parts, starting with adjusting the gate is often easier than changing the material—fiberglass aligns in the flow direction, and if the alignment is disturbed, the part will warp.
This also explains why we don't let customers switch to high fiberglass right away. Once the gate is changed and the mold temperature is raised, the flatness of the same batch of material may directly come back; adjusting the material first and then the mold is equivalent to shifting the process problem onto the material.
5. Four Common Misjudgments and the Real Causes
Defect 1: Warping means the material is soft, change to higher glass fiber.
This is the most regrettable category. When clients see warping in the frame, their first reaction is 'add fiberglass, switch to high-rigidity material.' The root cause of warping often lies in anisotropic shrinkage—fiberglass aligns along the flow direction, shrinking less in the perpendicular direction. For long board parts, one end shrinks more than the other, causing warping. Adjusting the gate location, mold temperature, or making a low-warp blend is often more effective than adding fiberglass. We have seen cases with the same material and mold, where simply changing the gate restored the flatness immediately.
Judgment 2: Use dry-state dimensions for positioning design.
Nylon absorbs moisture and swells. PA66 absorbs 8% moisture, so a 200 mm long locating edge can change by about 1 mm. If the pin holes are sized based on dry dimensions, after running in a high-humidity environment for six months, the positioning will be off. Dimensional reports must be based on conditioned (moisture-equilibrated) state, and positioning tolerances should be left according to the conditioned state.
Counter three: Ignoring the thermal expansion difference between inserts and cold plates.
The coefficient of linear expansion of aluminum is about 23×10⁻⁶/K, while PA66 is about 70–80×10⁻⁶/K, more than three times higher. Under temperature cycling, the interfacial stress accumulates, and cracks appear first around the insert. This is not due to insufficient material strength, but a mismatch in thermal expansion. The solution relies on interlocking structures and selecting a substrate with a similar CTE, which needs to be determined early.
Counterargument four: Once the flame retardant is added, that's it.
Halogen-free flame retardant systems require large amounts, and some flame retardants or antioxidants may migrate to the surface under long-term thermal cycling, resulting in precipitation, clouding, or yellowing. This is a pitfall on the formulation side—if the processing temperature exceeds the withstand temperature, or if the additives cannot tolerate the temperature, the surface becomes dirty. When you see yellowing or precipitation, first check the temperature tolerance of the additives before rushing to change the substrate.
A timeline (common insert cracking path in the industry): frame injection molding, room temperature assembly qualified → heat cycling during data center operation → stress accumulation around the insert → micro-cracks appear after six months to a year → expansion due to vibration during some handling → tracing back, it is due to CTE difference plus preloading. The problem was already embedded at the material selection stage, just slow to manifest.
When we receive an inquiry for a cold plate frame, we usually follow up with three questions: First, is the positioning accuracy based on dry-state or wet-state drawings; second, which metal cold plate is it paired with, and has the thermal expansion difference been calculated; third, is the annual usage enough to justify spreading the mold and verification costs. If these three questions cannot be clearly answered, the material is deemed unstable.
6. Processing and Verification: Warping and inserts should be monitored separately
The cold plate structural parts are injection-molded parts, with the two cavities controlled separately.
Drying. Nylon must be dried. If the moisture content of PA66 exceeds 0.15%, injection molding may cause hydrolytic degradation, resulting in reduced strength and high internal stress. Drying should be based on the actual measured moisture content to determine the window. In the rainy season in the south, a dehumidifying dryer is especially necessary.
Gate and runner. The gate position of large flat parts determines the flow balance and directly affects warpage. Discuss the gate first, then talk about material change.
Mold temperature. When the mold temperature is low, surface fiber floating, internal stress, and warping increase. For glass fiber parts, the mold temperature usually needs to be raised to the range of 110–120℃.
This 110–120℃ looks high, but it is the normal range for glass fiber parts; if the mold temperature is 20℃ lower, fiber floating and warping may significantly worsen. Mold temperature is not 'the higher the better,' it is 'enough for the glass fiber to be coated with resin and for the internal stress to be low'.
Insert. The temperature difference when preheating the metal insert, the interlocking structure, and the injection molding sequence all affect the interface. Add local ribs around the insert to distribute the stress.
Verification order. It is recommended to arrange it like this:
1. Material level: flexural modulus, thermo-oxidative retention rate
2. Dimensional Level: Flatness in dry and wet states, positioning dimensions
3. Component level: Temperature cycling Pull re-test
4. Flame Retardant Grade: UL94 / GWIT
5. System level: Final inspection before assembling the whole machine
The order cannot be changed. If the previous item hasn't passed, moving on makes the subsequent data meaningless.
There is a commonly overlooked detail: preheating the inserts. If metal inserts are not preheated before molding, the large temperature difference with the nylon causes the interface to shrink and delaminate as soon as it cools; preheating them to near the mold temperature ensures a firm grip. If this step is skipped, the dimensions may pass initially, but the inserts will loosen after a thermal cycle.
7. Boundaries: When not to use modified nylon
This section might be more valuable than the previous one.
Firstly, the cold plate itself needs to conduct heat. That's the property of metals (copper/aluminum); nylon does not conduct heat, don't think about that.
Secondly, locations near heat sources where the long-term temperature exceeds 150℃. Ordinary PA66 systems are not sufficient; semi-aromatic or PPS types are required.
Third, the main load-bearing framework of super large loads. The modulus and creep of nylon are what they are, while metal in such positions is more stable.
Fourth, the annual usage is too small to justify spreading the cost of injection molds and validation. Precision parts with positioning holes require mold opening, gate optimization, and thermal cycling, which is not feasible for a few hundred units per year.
Fifth, it is required that zero precipitation is visible and close to the electrical equipment. For this type, the expected time for flame-retardant precipitation should be aligned in advance, and if necessary, a cleaner system should be used.
Listing these five points at the outset is not to discourage, but to save time. I've seen more than one project where the samples were perfectly fine, but mass production had to be rolled back due to warping or insert cracks—the cost of rollback was much higher than if it hadn't been done in the first place.
There’s also the economic aspect: for a precision frame with positioning holes, the costs of mold opening, gate optimization, and heating cycle verification are often in the tens of thousands at a single time investment. If you need a few hundred pieces per year, the cost per piece ends up being higher than the material cost. Before starting a project, calculating the required quantity first is more practical than choosing the material grade.
8. Material Change Risk List (What needs to be changed when switching from the original plan to fiberglass nylon cold plate structural parts)
| link; segment; part | What needs to be moved? | Points that are easy to overlook |
|---|
| Mold | The gate is repositioned according to flow balance, without using the old position. | Fiber orientation causes warping |
| Dry | Determine the window based on the measured moisture content | Moisture carried in by recycled materials |
| Material Temperature / Mold Temperature | The mold temperature of fiberglass parts reaches the level of 110–120℃ | Give only according to the recommended value by grade |
| Pressure Holding and Demolding | Reinforce around the insert, set the order | Insert preheating temperature difference |
| Humidity control | Dimensions are measured in the wet state, not the dry state | Normal when dry, slightly abnormal when wet |
| Color difference | Advance confirmation of exterior color samples | Expected yellowing of flame-retardant system |
| Verification order | Material → Size → Insert → Flame Retardant → System | If the previous item fails, just move on. |
9. One-page report form (for those who need to report upward)
`
Project: Cold Plate Frame / Liquid Cold Plate Structural Components · Material Route Evaluation
Conclusion direction: Glass fiber nylon can be considered as a candidate, but whether it can be implemented depends on three prerequisites.
1. Three Rules That Must Be Followed
1. Flatness is determined in the wet state, not in the dry state.
2. Early matching of insert thermal expansion differences, without relying on material strength to force it
3. Address warp by activating the gate first, then discuss material changes
2. Precondition (It is recommended to postpone if any are not met)
· Long-term operating temperature ≤ 80℃ (locally verified separately)
· Solution matched for thermal expansion with cold-rolled metal sheets
· Annual usage is sufficient to offset the investment in injection molding and validation
3. Next Steps
1. Perform flatness and positioning dimension measurements in both wet and dry states
2. Perform insert temperature cycling and pull-out
3. Fixed Gate and Runner Balance Scheme
Risk warning: The main uncertainties of this route lie in warping and moisture-induced dimensions, not in initial strength.
`
10. Two Questions Frequently Asked by Readers
Question: How is it different from imported fiberglass nylon?
Only talk about two comparable things: for the same indicator, see whether it is marked with the test state (dry or wet); for the same part, see whether it provides long-term data on warpage and dimensional stability. Structural parts are extremely sensitive to state, so numbers with unknown states should not be compared directly. Some brackets have matured using domestic PA66-GF, while some precision positioning parts are still recommended to use semi-aromatic – regarding your flatness requirements, both temperature and humidity need to be considered.
Q: Is PA66-GF30 sufficient, or should we switch to a partially aromatic type?
Look at the position and precision. For purely supportive brackets with loose flatness requirements, PA66-GF30 is sufficient and cost-effective; for brackets with locating pin holes, attached to cold plates with thermal cycling, and requiring ten-year dimensional stability, the lower moisture absorption of PA6T-GF is more worthwhile. Before upgrading, do three things first: confirm that the wet-state dimensions are truly out of spec, confirm that creepage or structural margins have been optimized, and confirm that the temperature rise is within the semi-aromatic comfort zone. Only after completing these three steps should you upgrade, so that the money spent is well justified.
Make modified nylon: PA6, PA66, PA46, PA11, PA12, PA6T, PA9T, nylon alloys.
Modify PPO, PPS, and thermoplastic elastomers.
Produce nylon resin, secondary brand materials, and bulk materials for major chemical giants.
Additionally: We have long-term collection of nylon raw materials, sprue regrind, and various types of nylon waste, with proper disposal channels.
