你用同一个 FMEA 表去解释所有问题,就会在某个节点突然失语
这也是为什么新版 AIAG‑VDA 把 FMEA 做成七步法,先把结构和功能画清楚,再谈失效和风险.
风险与可靠性这块,不是会填表就行,关键在于你到底在分析哪一段失效链
产品没有按设计意图工作,或满足不了规定功能
没有按设计意图把东西做出来,偏差来自人机料法环
重点看监视能不能发现问题,系统响应能不能把后果压到安全状态
聚焦机器故障对质量和产能的影响,盯的是可靠性与可维护性
规划准备,结构分析,功能分析,失效分析,风险分析,优化,结果文件化
然后呢,表照样填得漂漂亮亮,可一到跨系统,跨软件,跨接口,就崩
我自己的感受是,七步法在提醒你两件事
先把结构画出来,再把功能讲清楚
结构不清,失效链一定会断
你得能回答,谁跟谁相连,谁影响谁
你得能回答,它存在的意义是什么,它要对谁负责
它不关心你现场有没有把螺丝拧紧
它关心的是,即便你把所有东西都按图纸做对了,这个设计本身会不会在某个工况下把你坑死
如果这个接口不稳定,会把哪条失效链点燃
最值钱的是你在结构分析和功能分析里画出来的那张网
那张网一旦画对了,失效链就不会乱飞.
它更像制造系统的自画像
你在里面要把每一步的交付物讲清楚,把每个工位可能制造的偏差讲清楚,把控制策略讲清楚
我用什么预防控制把它挡住,用什么检测控制把它抓出来
回到 PFMEA 这张图里去看,你会更明白他们在盯什么
他们盯的不是你写得好不好看,他们盯的是你有没有能力稳定地复制设计意图
可当产品进入客户手里,尤其是带软件和电子系统的复杂产品
真正决定风险上限的往往是两件事
第一件事,监视能不能发现故障
第二件事,发现之后系统怎么响应,能不能把它带回安全状态,或者至少把后果压下来
识别后会发生什么,报警,降级,限扭,切断,提示停车,还是静悄悄当没事
它的目的,是在客户运行期间帮助维持安全状态与合规状态,并关注监视与系统响应是否有效
而且它会把功能安全的思路带进来
产品的 PFMEA 写得再漂亮,设备一漂,夹具一磨,参数一跳,你依旧会报废
所以 MFMEA 看的不是产品怎么坏
它看的是设备怎么坏,以及设备坏的时候你有没有诊断与预警,有没有快速修复的设计
它最可能在什么部位退化,磨损,漂移
我能不能用诊断把它提前抓出来,让它别把整条线带崩
提高设备可靠性,缩短修复时间,并加入预防手段,比如诊断
这句话我很喜欢,因为它让 MFMEA 立刻从表格回到工程本身
Here is what happened.
At the customer site, 06:38
The head unit reports an abnormal sensor. Can the user keep driving?
Engineer
We already assessed this failure in our DFMEA, and the risk is not high.
Quality
What you assessed is the design failure chain. The customer is asking about monitoring and response during operation, so we need to look at FMEA-MSR.
Production
Stop arguing. Let's first compare the PFMEA and the control plan to decide how to contain this batch.
Equipment Engineering
One more thing: this line stopped three times yesterday. The root cause appears to be fixture wear, so we need to pull in the MFMEA.
With the same failure chain, from design to manufacturing to customer use, and then to equipment reliability itself, if you try to explain every problem with the same FMEA table, you will suddenly go silent at some node. This is why the new AIAG-VDA FMEA is built as a seven-step method: first map out the structure and functions clearly, then discuss failure and risk.
For risk and reliability, filling out a form is not enough. The key is identifying which segment of the failure chain you are actually analyzing.
First, spend 30 seconds telling the four brothers apart. Stop grabbing the wrong map.
DFMEA
Focuses on failures caused by design defects. The product does not work as designed, or cannot meet the specified functions.
What you are asking is whether the design has inherently built the risk in from the start.
PFMEA
Focuses on failures caused by the manufacturing process. The product is not built according to design intent, and deviations come from man, machine, material, method, measurement, and environment.
What you are asking is whether the process can stably replicate the design.
FMEA-MSR
The supplementary FMEA in the new manual focuses on whether monitoring can detect problems and whether system response can bring the consequences back to a safe state.
What you are asking is how the system will catch it when something goes wrong during customer use.
MFMEA
The FMEA for equipment and tooling focuses on the impact of machine failures on quality and capacity, targeting reliability and maintainability.
What you are asking is how equipment will fail, and whether you can know in advance before it fails.
In the new seven-step method, the real essence lies in steps two and three.
Many people hear "seven-step method" and start memorizing the sequence: planning and preparation, structure analysis, function analysis, failure analysis, risk analysis, optimization, and results documentation. Then what? The tables are still filled in beautifully, but when it comes to cross-system, cross-software, and cross-interface issues, it collapses. In my own experience, the seven-step method reminds you of two things: first, draw the structure out; second, explain the functions clearly. If the structure is unclear, the failure chain will definitely break.
What is structure analysis doing?
It breaks the system into systems, subsystems, components, and interfaces. You must be able to answer: who is connected to whom, and who affects whom.
What is function analysis doing?
It assigns to each structural element the function it should perform, as well as the inputs and outputs between it and its interfaces. You must be able to answer: what is its reason for existing, and who is it responsible to.
Note: The order of the seven-step method and the emphasis on "structure analysis" and "function analysis" can also be found in public interpretation materials for AIAG-VDA.
Back to DFMEA. What exactly are you preventing?
DFMEA is very much like writing a "self-judgment statement for the design." It does not care whether you tightened the screws on the shop floor. It cares about whether, even if you build everything exactly according to the drawings, the design itself could kill you under certain operating conditions.
A typical DFMEA question:
If this function becomes a little weaker, a little slower, or a little driftier, what happens? If this interface becomes unstable, which failure chain will it ignite?
You will find that the most valuable part of DFMEA is not the scoring. The most valuable part is the network you draw in structure analysis and function analysis. Once that network is drawn correctly, the failure chains will not fly around randomly.
Many failures do not occur inside a component, but at the interfaces.
Now look at PFMEA. It is really asking about replication capability.
PFMEA is not production's fault. It is more like a self-portrait of the manufacturing system. In it, you must clearly describe the deliverable of each step, the deviations that each station may create, and the control strategies.
A typical PFMEA question:
If this process deviates a little, is assembled wrongly a little, or misses inspection a little, which key characteristic will be thrown off? What prevention control do I use to block it, and what detection control do I use to catch it?
If you have long been under customer scrutiny about the control plan, go back and look at the PFMEA map, and you will understand better what they are scrutinizing. They are not watching whether you write nicely. They are watching whether you are capable of stably replicating the design intent.
The PFMEA result (risk analysis) directly determines the inspection frequency and methods in the control plan.
Many people, when they first hear about FMEA-MSR, think it sounds odd.
In the past, we talked about detection in DFMEA. But when the product enters the customer's hands, especially complex products with software and electronic systems, what really determines the upper limit of risk is often two things. First, can monitoring detect the fault? Second, after detection, how does the system respond, and can it bring the situation back to a safe state, or at least contain the consequences?
A typical FMEA-MSR question:
When a failure occurs during customer use, can the system or user identify it? After identification, what happens: alarm, degradation, torque limiting, shutdown, stop prompt, or quietly acting as if nothing happened?
Public materials mention that FMEA-MSR is a supplementary FMEA introduced in the AIAG-VDA 2019 version. Its purpose is to help maintain a safe and compliant state during customer operation, and to focus on whether monitoring and system response are effective. It also brings in the functional safety mindset.
Finally, let's talk about MFMEA, the one most factories tend to overlook.
You will discover a harsh truth: no matter how well the product PFMEA is written, once the equipment drifts, the fixture wears, or the parameters jump, you will still scrap parts. So MFMEA is not about how the product fails. It is about how the equipment fails, and whether you have diagnostics and early warning when the equipment fails, and whether you have designs for quick repair.
A typical MFMEA question:
What is the key function this equipment must complete? Where is it most likely to degrade, wear, or drift? Can I use diagnostics to catch it early, so it does not take the whole line down?
Some public materials state the goal of MFMEA very plainly: improve equipment reliability, shorten repair time, and add preventive measures such as diagnostics. I like this statement because it immediately brings MFMEA back from a table to engineering itself.
If you can take away only one page, take away these four sentences.
DFMEA: don't rush to score. First draw the structure and functions correctly.
If you draw the network wrong, everything after is ineffective effort.
PFMEA: focus on reproducibility, and treat the control plan as a commitment on the ground.
What you write, you must be able to do; what you do, you must be able to prove.
FMEA-MSR: treat the customer operation phase as a new battlefield. Monitoring and response are the key guardrails.
Making a mistake is not scary. What is scary is that the system does not catch it.
MFMEA: equipment degradation will eat all your yield. Move diagnostics and maintenance design forward.
Don't wait until the line stops to start writing lessons learned.
Original link: https://mp.weixin.qq.com/s/r2sO3k1Zan_nvKqzlFK5rg
