PFMEA Occurrence Rating

Quick reference

• 00:04 Hi, I'm Ray Sheen, I would now like to explain how to score probability of
• 00:09 occurrence for the Process FMEA.
• 00:11 This is still part of step five of our process for preparing a Process FMEA.
• 00:18 What we are evaluating is the likelihood that the failure will occur at some time
• 00:23 while operating the process for this product.
• 00:26 To determine the likelihood,
• 00:28 we first have to understand what would cause that failure.
• 00:31 So we list the possible causes, there may be more than one cause for
• 00:35 a particular failure mode.
• 00:37 Now based upon your experience, or whatever other relevant data you can find,
• 00:41 determine the frequency with which that cause occurs.
• 00:45 Or at least, the frequency of that cause in the intended process,
• 00:48 if you are doing this while creating a new process.
• 00:51 Evaluate each of the causes, independently.
• 00:54 So if you have a frequency data for a failure that could have multiple causes,
• 00:58 assign a portion of those failures to each cause.
• 01:02 But don't assign the entire proportion to all the causes.
• 01:05 That would be double counting the likelihood.
• 01:08 Let's look at the data sources for the frequency of the probability data.
• 01:12 If the process is already operating,
• 01:14 use actual defect data history from the process.
• 01:18 This is clearly the most accurate.
• 01:20 If the process is not yet
• 01:21 operational, use internal defect history data from similar processes.
• 01:26 You might need to go to several different processes to get workstation level data.
• 01:30 If there is no internal data,
• 01:32 then use external data sources such as supplier data or industry benchmark data.
• 01:38 Okay, now let's look at the actual scoring.
• 01:40 This scoring is much more numerically objective,
• 01:43 the only problem is that you may have very limited data to work with.
• 01:47 So I still provide the subjective rating scale with the objective numerical one.
• 01:52 Frankly, if it is a new process with no data, even though you don't
• 01:55 think it will happen, I would recommend a minimum of a 6,
• 01:58 unless you have a Poka Yoke to the process, so you can drop it down to a level 1.
• 02:03 So let's look at the scores.
• 02:05 Level 10 is very high, it is inevitable,
• 02:08 the data shows that it occurs more than 10% of the time.
• 02:12 Level 9 is still bad, it's very high and almost inevitable.
• 02:16 However, the data shows that it occurs at a rate of 1 in 20, or at about 5%.
• 02:21 Level 8 is high repeated failures, this is a rate of 1 in 50 or 2% of the time.
• 02:28 Level 7 is also rated as high repeated failures in the standards.
• 02:32 However, the rating has dropped to a 1 and 100, or 1% of the time.
• 02:38 Level 6 is described as moderate with occasional failures,
• 02:42 the rate is 1 in 500, or two-tenths of 1%.
• 02:45 For those of you who are familiar with Lean Six Sigma,
• 02:49 this is the score for process that has a CPK of 1.
• 02:52 Recall that I said this was my recommended score for any process, for
• 02:56 which you have no data.
• 02:58 You could select the higher score, but
• 03:00 I would not select anything lower until there was substantial data to prove it.
• 03:05 Level 5 is also described as a moderate, occasional failure.
• 03:09 The rate is 1 in 2,000, which means a .05% yield.
• 03:16 As we get to these lower numbers, it becomes harder sometimes to determine
• 03:20 what value to use, especially early in production.
• 03:23 Frankly, you shouldn't be selecting these unless you have a significant amount
• 03:27 of production data to support them.
• 03:29 Level 4 is still labelled as a moderate with occasional failure.
• 03:33 We're now talking about 1 failure in 10,000 opportunities.
• 03:37 This is just 100th of 1%.
• 03:40 Level 3 is labelled low, relatively few failures.
• 03:44 The ratio is 1 in 100,000, which means it is 1,000th of 1%.
• 03:50 And level 2 is is also low, with relatively few failures.
• 03:55 The ratio is less than 1 in a million,
• 03:57 and that's not just a saying, this is supported by data.
• 04:01 For the record, that is 1 10,000th of 1%.
• 04:05 Finally, level one,
• 04:06 which is labeled remote, very unlikely, but really means impossible.
• 04:11 The process has been Poka Yoke-ed, which means that there are design features in
• 04:15 the product or process that prevent the failure mode from occurring.
• 04:20 So let's wrap this up with a look at the FMEA, and the scoring for
• 04:23 the probability of occurrence.
• 04:25 As you can see, we have entered at least one cause for each failure mode, and
• 04:29 some failure modes have multiple causes.
• 04:32 The first cause, operator error, actually occurs rather frequently.
• 04:36 If you have real data for each step, use that.
• 04:39 In the absence of real data, I generally put that type of failure at a 6 or higher.
• 04:44 Quite frankly, that represents a well-trained set of operators,
• 04:48 who are performing at a high degree.
• 04:50 There are some studies that would suggest that the typical person makes a mistake
• 04:55 somewhere between 2% and 5% of the time.
• 04:58 If that is more representative of your people, you should be using a score of 8.
• 05:03 But for this example, I used 6, with all the operator error causes.
• 05:08 And I did the same for inspector error causes too.
• 05:12 The stockroom error was just 4.
• 05:13 There is some automation there, and
• 05:15 the historical data we had would support that value.
• 05:19 The wrong solution is also a 4.
• 05:22 There's only one solution at the workstation, so
• 05:25 that should minimize the failure.
• 05:27 But data shows that it does occasionally happen.
• 05:29 There is an automated decal selection and
• 05:32 inspection system in the stockroom, that checks the accuracy of each
• 05:35 decal as it is creating the page of decals needed for the lot.
• 05:39 The data from that system is really good, so we got a 3 there.
• 05:43 All the rest of the causes were based upon data except the last two near the bottom
• 05:47 of this slide.
• 05:48 Notice that the cause No Standard was set at a 1.
• 05:50 That means that it was Poka Yoke-ed, or mistake-proofed.
• 05:54 The way this was done is that the relevant portion of the standard along with
• 05:58 visuals, are printed in the work order, so the standard is always there.
• 06:03 Scoring failure modes for probability of occurrence means that you first must
• 06:08 think through what could realistically cause the failure mode to occur.
• 06:12 Then, look at the probability of that cause.

Lesson notes are only available for subscribers.