上个月有个做协作机器人关节的客户来问刚轮。
他没问"能不能做塑料"，问得更具体：PA12-CF30 这条路线，有没有人真做成了。
问完他发来一张截图，是某份公开资料里的一段话，图有点糊，只有一行数字看得清——0.5 μm / 300 mm。他说他就是被这行数字引过来的。
我把他那张截图放大看了两遍，回了他一句：这行数字是结果，不是条件。你现在拿它当条件去选料，找不到答案。
这篇就讲清两件事：这条路线的完整逻辑是什么，以及什么条件下它根本不成立。
一、刚轮的问题，从来不是"强度够不够"
先说清刚轮在谐波减速器里的位置。
谐波减速器三大件：波发生器、柔轮、刚轮。波发生器把柔轮撑成椭圆形，柔轮外齿于是在长轴两端和刚轮内齿啮上，短轴方向脱开。柔轮比刚轮少两个齿，波发生器转一圈，啮合位置就退两个齿——减速比就是这么来的。
刚轮在这里的角色，是固定不动的那个基准。它承担的是反作用扭矩，同时提供全部啮合齿的定位。
这个角色带来一个后果：刚轮的失效，绝大多数不是被压坏，是"位置变了"。
具体是三个地方在动：
齿隙。刚轮和柔轮的啮合侧隙，原本是按微米给的。齿隙一变，回差就变，机械臂末端的位置精度跟着变。
齿形。齿面磨下去一点点，接触区就变了，从面接触滑向边缘接触，然后磨损加速。
圆度与同轴度。刚轮是环形件，注塑的圆度一旦超差，装配后就是偏心运转——噪音先来，精度随后走。
所以问"塑料能不能做刚轮"，问法本身偏了。该问的是：在它的工作温度和寿命内，这个材料的齿形能不能守住。
刚轮的选材判据里，尺寸稳定排在强度前面。这句话在关节件上已经不是新观点，但在刚轮这个件上，它的权重还要再往上抬一档。
二、为什么路线会落到 PA12-CF30 上
从金属换到塑料，起点不是挑牌号，是搞清"金属为什么能用"。
钢制刚轮靠三件事吃饭：高模量（齿不变形）、尺寸不随环境变、齿面耐磨。 换成塑料，前两件都成了问题。
塑料的模量比钢低一到两个数量级。补偿的办法是加纤维——这是玻纤和碳纤增强的意义。但加水这件事只解决了刚度，解决不了吸水。
这就引出了这条路线的第一个关键选择：为什么是长碳链的 PA12，而不是更常见的 PA66。
| 体系 | 饱和吸水率（公开资料典型量级） | 熔点 | 加 30% 纤维后的特点 |
|---|
| PA66 + GF30 | 约 8–9% | 约 265℃ | 刚性好、耐温够，但吸湿尺寸漂移大 |
| PA46 + GF30 | 比 PA66 更高 | 约 295℃ | 流动性好、耐温更高，吸湿更敏感 |
| PA6 + GF30 | 约 8–10% | 约 220℃ | 成本低、工艺成熟，尺寸稳定性最弱 |
| PA12 + CF30 | 约 1.5% 量级 | 约 178–180℃ | 吸湿极小、韧性好；耐温上限低、成本高 |
看这张表，重点不在"哪个更好"，在差在哪。
PA66 的吸水率是 PA12 的五倍上下。落到一个齿顶圆直径几十毫米的环件上，吸湿饱和后的尺寸变化量级，就足以吃掉整个啮合侧隙的设计余量。
这不是"精度差一点"，是装配时的侧隙和运行半年后的侧隙不是同一个值。
低吸水到底重要到什么程度，这里说一个我们自己踩过的场景。
早年给客户做过一批精密环件，材料是 PA66-GF30，下线检测全部合格，报告按干态数据出。客户装了两周反馈噪音上来，拆下来复测尺寸超差。我们查了三天模具和注塑参数，什么都没查出来。
后来才想明白：件下线是干的，客户那边在南方雨季里放了半个月，件自己吸湿涨了。
补做调湿之后复测，尺寸正好落回公差带中间。件从头到尾都是好的，只是我们给的报告写的是错的状态。
那次之后，我们给精密件的交付流程加了一栏：报给客户的尺寸，必须是调湿之后测出来的那一组。干态数据留在内部，标明是过程记录，不随件出门。
这条规矩，就是刚轮这类件选 PA12 而不是 PA66 的原因。 不是 PA66 不够强，是它随环境变化的幅度，在刚轮上没法用补偿解决。
碳纤这一头还有两条要注意的：
- 碳纤的长度在注塑后会衰减，样条的模量数据不等于件上的模量，尤其是薄壁齿形区；
- 碳纤是导电的，靠近电机、编码器、线束的位置要考虑绝缘隔离，这一点在设计评审时就要提出来，不能等到 EMC 测试才发现。
一句话：刚轮选 PA12-CF30，买的是"不随环境变"这件事，不是买强度。强度是碳纤附赠的。
三、工况六维：刚轮到底被什么约束
刚轮的工况，比一般结构件复杂，因为它同时受机械和环境两条线夹击。
温度。 关节模组里，电机在近端，减速器在远端，刚轮的实际工作温度取决于散热路径。常见区间是 40–80℃，连续高负载或散热设计不足时会更高。要盯的是长期温度，不是峰值。
接触应力。 谐波传动的同时啮合齿数比例不高，齿面接触应力集中。这是刚轮与一般齿轮最大的区别——它不是均布受力。
摩擦与磨耗。 齿面在脂润滑条件下工作，摩擦系数和磨耗率决定寿命。这里有个容易忽略的点：对偶件是钢制柔轮，所以这是"塑料对钢"的摩擦副，不是塑料对塑料。 两者的选材逻辑完全不同。
介质。 润滑脂的相容性是硬条件。脂里的某些成分会让塑料件表面析出、发黏，甚至影响齿面尺寸。这一条在样机阶段很少测，到批量才暴露。
寿命与精度保持。 关节的往复次数是百万次量级。判据不是"什么时候断"，是"到寿命终点时齿形误差增加了多少"。
合规与洁净。 部分应用场景（洁净室、医疗、食品周边）对析出和挥发有额外要求，需要提前确认适用的标准语境。
把这六维摆在一起，会看到一个结论：刚轮没有一项指标是可以单独达标的，它们是耦合的。 温度高一点，模量掉一点，齿变形大一点，接触区就变；接触区一变，磨耗加速；磨耗一加速，齿隙就守不住。
四、选型判据表（这是本篇最该收藏的一页）
把上面的约束落成可核对的指标。下表的门限值是方向性建议，不是验收标准——实际数值必须由具体项目、具体工况和实测确定。
| 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 |
|---|
| 干湿态尺寸差 | 齿顶圆差异控制在 0.05% 以内 | 调湿前后实测 / ISO 294 | 齿隙漂移、装配后回差变大 | 选低吸水基材 + 强制调湿 | 材料本征决定，不靠助剂 |
| 齿面接触疲劳 | 10⁷ 次循环后齿形误差增量可控 | 关节台架 + 齿形测量 | 点蚀、齿面剥落 | 纤维增强 + 界面结合 | 偶联剂（纤维 / 树脂界面） |
| 摩擦系数（对钢，脂润滑） | 0.15 上下 | ASTM D1894 / SRV 试验 | 齿面磨损、局部温升 | 自润滑体系 + 对偶件配合 | 润滑剂 / 耐磨填料 |
| 长期热氧保留率 | 80℃×1000h 后 ≥75% | ISO 527 | 齿面发白、脆化 | 稳定化体系 | 抗氧剂（受阻酚 + 亚磷酸酯复配） |
| 圆度 / 同轴度 | 圆度按件精度定，通常 0.02 mm 量级起 | 三坐标 / 圆度仪 | 偏心、噪音、局部偏载 | 浇口与取向设计 | — |
| 弯曲模量 | 参考 8–10 GPa 量级（有取向依赖） | ISO 178 | 齿变形、啮合偏移 | 碳纤增强 + 结构补强 | 偶联剂（界面） |
| 润滑脂相容性 | 浸脂后尺寸与外观无异常 | 供应商脂样浸泡实测 | 表面析出、发黏、尺寸变化 | 提前确认脂体系 | 助剂迁移性需一并评估 |
怎么用这张表：不要逐行打分，先看第一行和第二行。这两行过不去，后面的都不用谈。
因为刚轮的失效逻辑是串联的——尺寸守不住，疲劳和磨耗的数据就失去意义。
一个提醒：表里"验证方法"一列，很多项目没有现成的国标可依（谐波传动的刚轮尤其如此）。没有标准可依时，就把验证方案写进技术协议，而不是省掉这一项。
五、四条常见失效，和它们的真实根因
失效一：齿隙在头两个月变大，之后稳定。
根因通常不是磨耗，是吸湿。件在出厂状态（干态或部分调湿）装上去，运行期间继续吸湿到平衡，尺寸往一个方向走。两个月后吸湿接近平衡，变化就停了。
通行解法：调湿态交付 + 复测。这一条比换料有效得多。
失效二：齿面局部发白，随后掉粉。
根因是热氧老化，而且往往是局部的——装配偏载导致某几个齿长期高温。这时候换更高的抗氧剂档位有用，但根因在装配同轴度，不在料。
失效三：齿面磨下去很快，但件本身没坏。
查两件事：一是对偶件（钢柔轮）的表面粗糙度与硬度；二是润滑脂是否与材料相容。塑料对钢的摩擦副，磨损往往发生在塑料侧，但诱因在钢材侧。
失效四：同一批件黄得深浅不一。
这不是"料不稳定"。分散不均的可能性更大——抗氧剂或色母在混料阶段就没混匀。看到这个现象，先查混料工艺和母粒化，别急着换料。
这里有一条要直说的：刚轮件的失效排查，先怀疑状态与工艺，最后才怀疑材料。 这和普通结构件是反的。因为刚轮的精度量级太小，任何状态波动都会被放大成"料不行"。
六、加工与验证：几件必须提前定的事
干燥。 长碳链尼龙的吸水率低，不等于不用烘。料在包装里受潮、车间湿度高、回用料掺入，都会把水分带进去。 干燥窗口要按实际含水率定，不能照抄牌号推荐值。
齿形的收缩补偿。 齿形不是圆，各处的收缩不一样，碳纤件的取向差异会放大这件事。模具的齿形补偿要按件做，不能按材料手册的通用收缩率给。
取向与浇口。 碳纤件是各向异性的，流动方向和垂直方向的模量、收缩都不同。刚轮是环形件，浇口位置直接决定圆度和齿形的一致性。
验证顺序。 建议这样排：
1. 尺寸与圆度（调湿后测，干态只做过程记录）
2. 齿形精度与啮合侧隙（装到实际柔轮上测，不用替代件）
3. 脂相容性（浸泡后复测尺寸与外观）
4. 关节台架（跑疲劳，中途复测齿形）
5. 环境叠加（温度循环 + 湿度循环，最后再上整机）
顺序不能换。 前一项不通过就往下走，后面测出来的数据没有解释意义。
这里有个内行细节：刚轮件的尺寸，注塑下线后 24 小时测一次、调湿完成后测一次，两次数据的差，比绝对值更有用。 差值大说明这个件对状态敏感，装配环境的湿度必须写进协议。
七、边界：什么时候这件事根本不该谈
这一段可能比前面六段更值钱。
以下四种情况，刚轮走塑化这条路不建议推进：
其一，长期工作温度超过 110℃。 PA12 体系在这个温度区间长期运行的性能保持数据，公开资料里的支撑不足。这不是改配方能解决的问题，换体系也难——因为低吸水和耐高温在尼龙家族里本来就是一对矛盾。
其二，要求刚轮承担主承力路径，且精度等级很高。 塑料的模量和蠕变特性，决定了它不适合作为高精度主承力件。这是材料物理层面的边界。
其三，整机的验证资源不足以支撑长周期台架。 刚轮的塑化验证不是一次样品测试能收口的，要跑到百万次量级、期间多次复测齿形。没有这个验证预算，就不要开这个头。
其四，用量小到无法摊薄模具与验证成本。 这个件要开专用模具、做齿形补偿、走长周期验证。如果年用量只有几百件，从钱上就不成立。
把这四条写在前面，不是劝退，是省时间。 我见过太多项目在样品阶段很顺，卡在批量验证上，最后整个方案回退——回退的成本，比一开始就不做高得多。
还有一个必须说清的：柔轮是另一回事。 柔轮靠弹性变形传扭，超千万次疲劳叠加微米级形变控制，这超出了热塑性材料目前能稳定支撑的区间。刚轮能谈，柔轮先不谈——这两件不能打包一起做决定。
换料风险清单（从金属换到 PA12-CF30，要动的东西）
| 环节 | 要动什么 | 容易漏的点 |
|---|
| 模具 | 齿形按件做收缩补偿，不能套通用收缩率 | 碳纤取向导致的方向性收缩差 |
| 干燥 | 按实测含水率定窗口，不是照抄推荐值 | 回用料掺入带入的水分 |
| 调湿 | 强制调湿 + 称重判定 + 复测尺寸 | 按平均壁厚估时间，厚壁处没吸透 |
| 料温 / 模温 | 按齿形填充与圆度要求联合调 | 只按牌号推荐值给，不看件 |
| 保压与脱模 | 环形件易变形，保压曲线与脱模方式要重定 | 沿用原金属件的工装思路 |
| 色差 | 碳纤件本身颜色深且批次有差 | 外观件的色差标准要提前放宽或另定 |
| 验证顺序 | 尺寸 → 齿形 → 脂相容 → 台架 → 环境叠加 | 前一项未过就往下走 |
一页纸汇报表（给要向上汇报的人）
`
项目：谐波减速器刚轮 · 材料路线评估
结论方向：PA12-CF30 可作为候选路线，能否落地取决于四项前置条件
一、必须守住的三条
1. 调湿态交付，干态数据不上报告
2. 齿形按件做收缩补偿，不套通用值
3. 全程无润滑脂相容性数据不进入台架
二、前置条件（任一不满足则建议暂缓）
· 长期工作温度 ≤ 110℃ 量级
· 有百万次量级台架验证预算与周期
· 年用量足以摊薄模具与验证成本
· 装配同轴度可控（偏载会放大齿面老化）
三、下一步动作
1. 取柔轮实物，测实际啮合侧隙
2. 做调湿前后尺寸差，评估件对状态的敏感度
3. 脂相容性浸泡试验（两周起）
风险提示：本路线的主要不确定性在长期精度保持，不在初始强度。
`
读者常问的两句
问：既然 PA12 耐温只有 180℃ 熔点，那关节模组里靠近电机的刚轮是不是就没戏？
要看实际温度，不要看电机规格。关节模组的散热路径决定了刚轮的真实温度，很多时候比想象中低。做法是实测——在样机上贴热电偶，跑典型工况，取长时间稳态值。用峰值温度判断会误杀方案，用规格书温度判断会误放。
问：碳纤和玻纤，这个件能不能换？
方向不同。碳纤给的是模量和尺寸稳定性，玻纤给的是成本和韧性。刚轮这类对模量和翘曲都敏感的环形件，碳纤的理由更充分；但碳纤的导电性和成本必须一起考虑。另外换纤维体系等于换一套界面方案，验证要重做，不能只对比样条数据。
结语
谐波减速器刚轮的塑化，说到底是一道精度保持题，不是强度题。
判断链只有三条：
温度定体系 → 吸水率定牌号 → 验证顺序定成败。
三条都定完，"能不能用塑料"这个问题自然就有答案了。
如果你手上正有一个刚轮或关节件要定料，把三样东西发过来就能给方向：长期工作温度、实际啮合侧隙要求、年用量量级。
站在树脂厂和注塑厂之间，很多件其实已经定了一半——定它的是工况、是精度、是验证顺序，不是牌号。
我们做的事很具体：把 PA6、PA66、PA46、PA11、PA12、PA6T、PA9T 和尼龙合金这些树脂，改成某个件真正能用的样子；顺带做改性 PPO、PPS 和热塑性弹性体。
也经营各大化工巨头的尼龙树脂、副牌料和大包料，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。
刚轮这类件的选料与试模，可以一起聊。
Last month, a customer who makes collaborative robot joints came to inquire about geared pulleys.
He didn't ask 'whether it can be made with plastic,' he asked more specifically: for the PA12-CF30 route, has anyone actually made it?
After asking him, he sent a screenshot, which was a passage from some public information. The picture was a bit blurry, and only one line of numbers was clearly visible — 0.5 μm / 300 mm. He said he was drawn in by this line of numbers.
I zoomed in on his screenshot and looked at it twice, then replied to him: These numbers are the results, not the conditions. If you use them as conditions to select materials, you won't find the answer.
This article explains two things clearly: what the complete logic of this route is, and under what conditions it simply does not hold.
1. The problem with solid wheels has never been whether the "strength is enough".
First, clarify the position of the sun gear in the harmonic reducer.
The three main components of a harmonic reducer are: the wave generator, the flexspline, and the circular spline. The wave generator deforms the flexspline into an elliptical shape, causing the outer teeth of the flexspline to engage with the inner teeth of the circular spline at the ends of the long axis, and disengage along the short axis. The flexspline has two fewer teeth than the circular spline, so for each complete rotation of the wave generator, the engagement position moves back by two teeth — this is how the reduction ratio is achieved.
The role of the sun gear here is the fixed reference. It bears the reaction torque while providing the positioning for all the meshing teeth.
This role brings a consequence: the failure of the just-turned wheel is mostly not due to being crushed, but because its 'position has changed'.
Specifically, there are three places that are moving:
Backlash. The backlash between the drive wheel and the driven wheel was originally given in microns. When the backlash changes, the hysteresis changes, and the position accuracy of the robotic arm's end effector changes accordingly.
Tooth shape. When the tooth surface is worn down a little, the contact area changes, shifting from surface contact to edge contact, and then the wear accelerates.
Roundness and coaxiality. The gear wheel is a ring-shaped part. Once the roundness of the injection molded part exceeds the tolerance, it will run eccentrically after assembly — noise comes first, followed by precision.
So asking 'Can plastic be used to make gears?' is misguided. The question should be: Within its operating temperature and lifespan, can this material maintain the shape of its teeth?
In the selection criteria for the gear wheel, dimensional stability ranks ahead of strength. This idea is no longer new for joint parts, but for the gear wheel, its importance needs to be elevated even further.
2. Why the route ends up on the PA12-CF30
When switching from metal to plastic, the starting point is not choosing the grade, but understanding why 'metal can be used'.
Steel gear wheels rely on three things to function: high modulus (teeth do not deform), dimensions do not change with the environment, and tooth surfaces are wear-resistant. Changing to plastic causes the first two issues.
The modulus of plastic is one to two orders of magnitude lower than that of steel. The way to compensate is to add fibers — this is the purpose of glass fiber and carbon fiber reinforcement. But adding water only solves the stiffness problem; it does not solve water absorption.
This leads to the first key choice of this route: why it is long-chain PA12 instead of the more common PA66.
| system | Saturated water absorption rate (typical magnitude in public data) | Melting point | Characteristics after adding 30% fiber |
|---|
| PA66 GF30 | About 8–9% | About 265℃ | Good rigidity and sufficient temperature resistance, but large dimensional drift due to moisture absorption |
| PA46 GF30 | Higher than PA66 | approximately 295°C | Good liquidity, higher temperature resistance, more sensitive to moisture |
| PA6 GF30 | About 8–10% | About 220℃ | Low cost, mature process, weakest dimensional stability |
| PA12 CF30 | About 1.5% level | About 178–180°C | Very low moisture absorption, good toughness; low upper temperature limit, high cost |
Looking at this table, the focus is not on 'which is better,' but on where the differences lie.
The water absorption rate of PA66 is about five times that of PA12. When applied to a ring component with a tip circle diameter of several tens of millimeters, the dimensional change after moisture saturation is enough to consume the entire design margin of the meshing clearance.
This is not 'slightly less accurate'; the clearance during assembly and the clearance after six months of operation are not the same value.
How important low water absorption really is—here, I'll talk about a scenario we've personally experienced.
In the early years, we made a batch of precision rings for a client, using PA66-GF30 as the material. All off-line tests passed, and the report was based on dry-state data. After the client installed them for two weeks, they reported noise issues, and upon disassembly and re-measurement, the dimensions were out of tolerance. We checked the mold and injection molding parameters for three days but found nothing.
Later I realized: the item was produced offline, and the client had kept it in the southern rainy season for half a month, so the item itself absorbed moisture and expanded.
After redoing the humidity adjustment and retesting, the dimensions just fell back in the middle of the tolerance range. The part was good from start to finish, it's just that the report we provided stated the wrong condition.
After that time, we added a column to the delivery process for precision parts: the dimensions reported to the customer must be those measured after humidity adjustment. The dry-state data remain internal, marked as process records, and do not leave with the parts.
This rule is the reason why parts like this gear are selected as PA12 instead of PA66. It's not that PA66 isn't strong enough, but the extent to which it changes with the environment cannot be compensated for on the gear.
There are also two things to pay attention to at this end of the carbon fiber:
- The length of carbon fibers will decrease after injection molding, and the modulus data of the spline is not equal to the modulus on the part, especially in the thin-walled tooth region;
- Carbon fiber is conductive, so when it is close to motors, encoders, or wiring harnesses, insulation and isolation must be considered. This should be raised during the design review and cannot be discovered only during EMC testing.
In a word: I just chose PA12-CF30; I bought it for the fact that it "doesn't change with the environment," not for its strength. The strength is just a bonus from the carbon fiber.
3. Six-dimensional working condition: What exactly constrains the rigid wheel
The operating conditions of the newly machined wheel are more complex than those of ordinary structural parts because it is subjected to the dual pressures of both mechanical and environmental factors.
Temperature. In joint modules, the motor is at the proximal end and the reducer is at the distal end. The actual operating temperature of the rigid wheel depends on the heat dissipation path. The common range is 40–80℃, and it can be higher under continuous high load or insufficient heat dissipation design. The focus should be on the long-term temperature, not the peak value.
Contact stress. In harmonic transmission, the ratio of meshing teeth is not high, leading to concentrated contact stress on the tooth surface. This is the biggest difference between rigid wheels and general gears — the force is not uniformly distributed.
Friction and wear. The tooth surface operates under grease lubrication, and the friction coefficient and wear rate determine the lifespan. Here is a point that is easy to overlook: the mating part is a steel soft wheel, so this is a 'plastic-to-steel' friction pair, not plastic-to-plastic. The material selection logic for the two is completely different.
Medium. The compatibility of grease is a strict requirement. Certain components in the grease can cause the surface of plastic parts to exude, become sticky, or even affect the dimensions of the gear teeth. This aspect is rarely tested during the prototype stage and only becomes apparent during mass production.
Life span and accuracy retention. The number of reciprocating cycles of the joints is in the millions. The criterion is not 'when it breaks,' but 'how much the tooth shape error has increased by the end of its life.'
Compliance and cleanliness. Some application scenarios (cleanrooms, medical, food-related) have additional requirements for leaching and volatilization, and the applicable standard context needs to be confirmed in advance.
Putting these six dimensions together, a conclusion can be seen: no indicator of the gear wheel can meet the standard on its own; they are coupled. If the temperature is a little higher, the modulus drops a little, and the tooth deformation is a bit larger, the contact area changes; once the contact area changes, wear accelerates; once wear accelerates, the tooth clearance can no longer be maintained.
4. Selection Criteria Table (This is the page you should collect the most in this article)
Translate the above constraints into verifiable indicators. The threshold values in the table are directional suggestions, not acceptance criteria—the actual values must be determined by specific projects, specific conditions, and actual measurements.
| Indicator | Directional Threshold | Verification Method / Standard | Common Failures | Common solution | Corresponding auxiliary agent system |
|---|
| Difference in dimensions between dry and wet states | The difference in addendum diameter is controlled within 0.05% | Measured before and after humidity adjustment / ISO 294 | Tooth clearance drift, increased backlash after assembly | Choose low water-absorption substrate Forced humidity adjustment | Determined by the material's intrinsic properties, independent of additives |
| Tooth surface contact fatigue | The incremental gear error is controllable after 10⁷ cycles | Joint Test Bench Gear Measurement | Pitting, tooth surface flaking | Fiber Reinforcement Interface Bonding | Coupling Agent (Fiber / Resin Interface) |
| Coefficient of friction (for steel, grease lubrication) | 0.15 up and down | ASTM D1894 / SRV Test | Tooth surface wear, local temperature rise | Self-lubricating system Mating parts fit | Lubricant / Wear-resistant Filler |
| Long-term thermal-oxygen retention rate | After 80℃ × 1000h ≥75% | ISO 527 | Tooth surface whitening and brittleness | Stabilization system | Antioxidant (hindered phenol and phosphite blend) |
| Roundness / Coaxiality | Roundness is determined according to the precision per piece, usually starting at the 0.02 mm level. | CMM / Roundness Tester | Eccentricity, noise, local overload | Gate and Orientation Design | — |
| Flexural Modulus | Reference 8–10 GPa range (orientation-dependent) | ISO 178 | Tooth deformation, meshing deviation | Carbon fiber reinforced structural strengthening | Coupling agent (interface) |
| Lubricant Compatibility | No abnormalities in size and appearance after resin impregnation | Actual measurement of vendor grease sample soaking | Surface precipitation, stickiness, dimensional changes | Confirm the lipid system in advance | The migration of additives needs to be assessed as well |
How to use this table: Do not score line by line. First, look at the first and second lines. If these two lines are not passed, you don't need to discuss the rest.
Because the failure logic of the gear train is serial—if the dimensions are not maintained, the fatigue and wear data become meaningless.
A reminder: In the 'Verification Method' column of the table, many items do not have existing national standards to refer to (especially the rigid wheel of harmonic drives). When there is no standard to follow, the verification plan should be written into the technical agreement, rather than omitting this item.
Five, four common failures and their real root causes
Failure 1: The gear backlash increased in the first two months and then stabilized.
The root cause is usually not wear, but moisture absorption. The parts are installed in their factory state (dry or partially conditioned), and during operation they continue to absorb moisture until equilibrium is reached, causing the dimensions to change in one direction. After about two months, as moisture absorption approaches equilibrium, the change stops.
Common solution: Deliver in a humid state and retest. This is much more effective than changing the material.
Failure 2: The tooth surface becomes locally white, followed by powdering.
The root cause is thermal oxidative aging, and it is often localized—assembly misalignment causes certain teeth to be at high temperature for a long time. At this point, using a higher grade antioxidant can help, but the root cause lies in assembly coaxiality, not in the material.
Failure three: The tooth surface wears down quickly, but the part itself is not damaged.
Check two things: first, the surface roughness and hardness of the mating part (steel-flexible wheel); second, whether the grease is compatible with the material. In the friction pair of plastic against steel, wear often occurs on the plastic side, but the cause originates from the steel side.
Failure 4: The same batch of items has inconsistent yellowing depth.
This is not 'unstable material.' Uneven dispersion is more likely—the antioxidant or color masterbatch was not thoroughly mixed during the blending stage. When you see this phenomenon, first check the blending process and pelletizing, and don’t rush to change the material.
Here is something that needs to be said directly: for failure investigation of gear wheels, first suspect the condition and process, and only then suspect the material. This is the opposite of ordinary structural parts. Because the precision level of gear wheels is so small, any fluctuation in condition will be amplified into 'the material is no good'.
6. Processing and Verification: Several Things That Must Be Decided in Advance
Drying. The low moisture absorption of long-chain nylon does not mean it does not need to be dried. Moisture can be introduced when the material gets damp in packaging, the workshop has high humidity, or recycled material is mixed in. The drying window should be determined based on the actual moisture content, and the recommended values from the grade should not be copied directly.
Toothing shrinkage compensation. The teeth are not round, and the shrinkage varies in different areas. Differences in the orientation of carbon fiber parts will amplify this issue. The mold's tooth compensation needs to be done per part and cannot use the general shrinkage rate from the material manual.
Orientation and gate. Carbon fiber parts are anisotropic, with different modulus and shrinkage in the flow direction and the perpendicular direction. The flywheel is a ring-shaped part, and the gate location directly determines the roundness and tooth profile consistency.
Verification order. It is recommended to arrange it like this:
1. Size and roundness (measured after humidity adjustment, only recorded during the dry state process)
2. Tooth profile accuracy and backlash on the meshing side (measured on the actual flexible gear, no substitutes used)
3. Lipid Compatibility (Size and Appearance Re-measured After Soaking)
4. Joint test rig (run fatigue test, retest tooth profile midway)
5. Environmental stacking (temperature cycling, humidity cycling, and finally the whole machine)
The order cannot be changed. If the previous item fails, move on, the data measured afterwards has no explanatory significance.
Here's an insider detail: the dimensions of newly machined parts should be measured once 24 hours after injection molding, and once after humidity adjustment is completed. The difference between the two sets of data is more useful than the absolute values. A large difference indicates that the part is sensitive to conditions, and the assembly environment's humidity must be specified in the protocol.
VII. Boundaries: When This Should Never Be Discussed
This section might be more valuable than the previous six sections.
In the following four situations, it is not recommended for Ganglun to proceed with the path of plasticization:
First, the long-term working temperature exceeds 110°C. There is insufficient publicly available data on the performance retention of the PA12 system operating in this temperature range for a long time. This is not a problem that can be solved by changing the formula, and switching systems is also difficult—because low water absorption and high temperature resistance are inherently contradictory in the nylon family.
Second, it requires the rigid wheel to bear the main load path, with a very high level of precision. The modulus and creep characteristics of plastic determine that it is not suitable as a high-precision main load-bearing component. This is a limitation at the material physics level.
Third, the verification resources for the entire machine are not sufficient to support long-cycle test benches. The plastification verification of the flywheel cannot be completed with a single sample test; it requires running to the scale of millions of cycles, with multiple re-tests of the tooth profile in between. Without this verification budget, do not start this task.
Fourth, the quantity is too small to justify diluting the mold and validation costs. This part requires a dedicated mold, tooth profile compensation, and a long verification cycle. If the annual usage is only a few hundred pieces, it is financially unfeasible.
Writing these four points upfront is not to discourage, but to save time. I have seen too many projects proceed smoothly at the sample stage, only to get stuck in mass validation, and in the end, the entire plan regresses—the cost of regression is much higher than not doing it in the first place.
There is one more thing that must be clarified: a flexible wheel is a different matter. A flexible wheel transmits torque through elastic deformation, with over ten million cycles of fatigue combined with micrometer-level deformation control, which exceeds the range that thermoplastic materials can currently support stably. We can discuss rigid wheels, but let's not discuss flexible wheels for now — these two things cannot be decided together.
Material Change Risk List (from metal to PA12-CF30, things that need to be changed)
| link; segment; part | What needs to be moved? | Points that are easy to overlook |
|---|
| Mold | Tooth forms should have shrinkage compensation made per piece, and cannot use a general shrinkage rate. | Directional shrinkage difference caused by carbon fiber orientation |
| Dry | Set the window based on the actual measured moisture content, not by copying the recommended value. | Recycled materials mixed with the water content brought in |
| Humidity control | Forced humidity adjustment Weight-based determination Re-measure dimensions | Based on the average wall thickness to estimate the time, the thick-walled areas are not fully soaked. |
| Material Temperature / Mold Temperature | Joint adjustment according to tooth profile filling and roundness requirements | Only give according to the recommended value by grade, without looking at the pieces |
| Pressure Holding and Demolding | Ring-shaped parts are prone to deformation, and the pressure-holding curve and demolding method need to be redefined. | Follow the tooling approach of the original metal parts |
| Color difference | The carbon fiber parts themselves are dark in color and have batch variations. | The color difference standards for exterior parts should be relaxed in advance or set separately. |
| Verification order | Size → Tooth profile → Fat compatibility → Test bench → Environmental superposition | If the previous item fails, just move on. |
One-page report form (for people who need to report upwards)
`
Project: Harmonic Reducer Flexspline · Material Route Assessment
Conclusion direction: PA12-CF30 can be considered as a candidate route, and whether it can be implemented depends on four prerequisite conditions.
1. Three Rules That Must Be Followed
1. Deliver in a humidity-controlled state; dry-state data will not be reported
2. The gear shape is compensated for shrinkage individually, not using a general value.
3. Full-process data without grease compatibility will not enter the test stand
2. Precondition (It is recommended to postpone if any are not met)
· Long-term operating temperature ≤ 110℃ range
· Budget and schedule for test bench verification on the scale of millions of times
· Annual usage is sufficient to dilute mold and validation costs
· Assembly coaxiality is controllable (eccentric load will accelerate gear surface aging)
3. Next Steps
1. Take the actual flexible wheel and measure the actual meshing backlash
2. Measure the dimensional differences before and after humidity adjustment, and assess the sensitivity of the part to the condition
3. Lipid Compatibility Soaking Test (starting from two weeks)
Risk Warning: The main uncertainty of this route lies in long-term accuracy maintenance, not in initial strength.
`
Two questions readers often ask
Q: Since the PA12 temperature resistance is only 180°C melting point, does that mean the rigid wheel near the motor in the joint module is out of the question?
You need to look at the actual temperature, not the motor specifications. The heat dissipation path of the joint module determines the actual temperature of the solid wheel, which is often lower than expected. The approach is to measure it—attach a thermocouple to the prototype, run typical operating conditions, and take the long-term steady-state value. Using the peak temperature to judge can wrongly eliminate options, and using the temperature from the specifications can wrongly approve them.
Question: Can this part be replaced with carbon fiber or fiberglass?
Different directions. Carbon fiber provides modulus and dimensional stability, while fiberglass provides cost-effectiveness and toughness. For rigid rings such as wheels, which are sensitive to modulus and warping, the reason for using carbon fiber is more compelling; however, the conductivity and cost of carbon fiber must also be taken into account. Additionally, changing the fiber system is equivalent to changing the entire interface scheme, so verification must be redone, and cannot be done by simply comparing spline data.
Conclusion
The plasticization of the flexspline in a harmonic drive reducer, after all, is a matter of maintaining accuracy, not strength.
The judgment chain has only three links:
Temperature determines the system → Water absorption determines the grade → Verification sequence determines success or failure.
Once all three are set, the question of 'whether you can use plastic' naturally has an answer.
If you currently have a pin or joint component that needs material selection, sending over three things can provide direction: long-term operating temperature, actual required meshing clearance, and the annual quantity level.
Standing between the resin factory and the injection molding factory, many parts are actually already half decided—the deciding factors are the working conditions, the precision, and the verification sequence, not the grade.
What we do is very specific: we take resins like PA6, PA66, PA46, PA11, PA12, PA6T, PA9T, and nylon alloys, and turn them into a form that can actually be used for a certain part; we also do modified PPO, PPS, and thermoplastic elastomers along the way.
Also operates the nylon resin, secondary-grade materials, and bulk materials of major chemical giants, and has long-term purchasing of nylon raw materials, sprue regrind, and various types of nylon waste, with formal disposal channels.
The selection of materials and test molding for parts like just-turned wheels can be discussed together.
