What is CP(k)? Many people ask this question. I have a simple explanation which possible will help. Cp(k) is the capability of a process on the side of the process distribution closest to the risk (specification) Risk in this case is going outside the specification which presumably has dire circumstances or a bad effects.
The best way to describe this is by utilizing an example everyone can relate to. Lets talk about a road. The width of 1 lane will represent the blueprint specifications high and low. The width of the specification is set by the design engineer while preparing the design. This represents one half af the ratio required to calculate Cp(k). The other half is related to the variation of the car and its’ driver. No car can drive straight down the road without some variation. This variation is made up of the things we typically look at when constructing a Fishbone Diagram (Ishikawa). Man, Method, Machine, Material, Measurement, and Mother Earth(environment). These common cause variables determine the standard deviation that is calculated in order to construct the distribution.
Cp(k) then becomes the ratio of the distance from the center of where the car has been travelling in real time divided by half of the width of the bell curve (3 sigma or 3 standard deviations) This ratio is simply analysed by evaluation whether it is greater than 1 or not. Less than 1 means the car is going over the edge of the road on one side and if closer to the other side of the lane, crossing into oncoming traffic. The k factor is only expressed on the side that has the greatest risk…. 1.33 ie 4 sigma capable, which provides for some (1/3 more) margin of safety. 1.67 has a greater margin etc…
In order to do a good job, Cp(k) must have a sample size large enough to determine a behavior of the system (car and driver etc…). Too few and one can surmise that the process is capable when in fact we have only observed the car driving down one block. Applying what we see on one block does not provide for a comfortable extrapolation to a trip across the country. This is important as Cp(k) is best used when sample sizes are greater than 30 where the amount of sample error is reduced to an acceptable amount when considering the cost of measuring many more. Fewer than 30 and we are making a judgment that a driver and car are capable of driving across the country without leaving the lane by only observing how well they did while driving the relatively short distance of one block. A whole lot or risk there in my opinion.
The sequence of when to take an action is illustrated on our website. I have imported the picture below to explain to all of those who have asked about the order:
Can you determine the order of need for change in the following three examples?
- Severity (5), Occurrence (4), Detection (2) = 40
- Severity (9), Occurrence (2), Detection (2) = 36
- Severity (8), Occurrence (1), Detection (8) = 64
The correct order for action is #2, #1, #3. To find out why, please contact us or sign up for our value driven training on FMEA.
The order is Severity First (item 2) (9 and 10) without any occurrence or RPN - the purpose for this is to find a design or process change that eliminates the Failure Mode. This is very difficult, but possible; an example is Tire Blowout, where the mitigating action is to put a “run flat” tire on the car/truck.
Item 1 is next because of the Occurrence linked to the 5 in severity. Reduction in Occurrence is the second responsibility of the design or process team when looking at an FMEA. The RPN is not relevent. Poka Yoke, or reduction in variation or conversely tolerance relief will reduce the Occurrence. This relates to error proofing and if it is not feasible then capability improvement (CP index > or = to 1.67)
Finally the detection number is addressed, but generally only when a design or process has been determined to have a high enough occurrence and a high enough severity to be concerned about the risk. Target should be 3 or below.
If there any specific questions, just email me and I will address them in a future blog post.
Lee
I was returning a car the other day at Hertz at the Detroit airport. And lo and behold, on the wall upstairs were large poster sized examples of how to do an 8D. 8D, as you may know, is a great way to get to a root cause of a problem by working it from the Symptom to the Problem Statement using the “5 why” process. From there the statement is converted to a problem description using an IS/IS NOT analysis. The IS/IS NOT gathers facts about the what, where, when and How Big. This analysis is more about finding out what the problem is not as opposed to looking for the root cause initially.
The root cause is only agreed upon after developing theories based on the facts and the differences and changes that have taken place. Change is usually where the problem is. The fact that change takes place without proper verification and validation makes the 8D process very effective. The team does not have to be Six Sigma experts to use it effectively, but it ties in nicely with six sigma process.
I hope Hertz uses the process effectively and I wish them good luck and applaud the effort to find the root cause and fix it at that level.
Just another note on the 8D relationship to the FMEA, I am sitting in a facilitation of the Lean FMEA approach and one of the matrix items is the 8D inputs. The past failures is one item in the 3 cases for doing an FMEA.
The FMEA and more appropriately the Lean design Matrix for the FMEA family should always be present for the 8D development. The brainstormed causes already present in the FMEA can shorten the root cause activity by at least one and possibly two layers of the “WHY” discussions.
Keep in mind that the FMEA will be updated at D7 (prevention) or if done most efficiently in real time at D2 and D4 as the 8D progresses.
Lee
Just got an oppurtunity to look over the new handbook and am very pleased with the outcome. The inclusion of prework documents and clarification in the RPN useage (to not to for actions), clarification on Design FMEA detection ratings and some improvement in Severity explanation ( groups like 5 and 6, 7 and 8, and 9 and 10 will make it easier to select Severity initially.)
The most interesting item is the new forms in the back I am especially interested in the Form that had the design or process controls prevention side by side with the cause and then the Occurrence column. This is the most logical of all. The prevention controls impact occurrence so it should appear before the occurrence column as in the new formats.
Over all a good job and an excellent update. It appears the team at AIAG is doing the right thing for the future of FMEA. Bravo…
Lee
Back in the USA and teaching at one of the Japanese automaker’s design facilities in the Detroit Area today. The topic came up as to how the Test plan can be affected by the Design FMEA activity.
My response to the question was as follows:
One of the Design FMEA’s jobs is to determine that the testing and prevention methods are adequate to either prevent a failure mode and cause mechanism from happening, (Prevention affects Occurrence) and to determine if a test or groups of tests will excite a failure mode either directly or through a cause which was deemed probable during the analysis (This is the Detection number) A detection ranking is assigned to each test and the lowest number is placed in the Detection column. Here is where the process goes wrong.
The purpose of the the detection ranking is to give an honest assessment of the tests to determine if they are harsh enough and have the causes which have high Occurrence represented within it. If the test does not consider a cause that has a high Occurrence then the ranking should climb. A ranking greater than 3 (my opinion not to standard) then an action should be considered to improve the test to make it more representative of how the team invisions a failure. The Detection action (with no regard to RPN) is captured and given a timing and responsibility for follow-up. The RPN has nothing to do with determining if a test improvement is necessary.
The RPN is assigned to the action (not the other way around) and the resulting RPN after the action is taken is compared to the original for overall relative risk improvement. In fact the RPN is the only way to tell (numerically) if actions were successful in general. The RPN has each element where actions can be taken so it represents a good overall risk reduction comparison. The key word is comparison. OOOPs getting off topic here. Back to testing.
The test improvements are intended to excite the failure mode with the high occurrence causes present in order to fail the part or system, etc… On the other side the design engineer, knowing a risk exists takes an action to prevent a failure so the two opposing forces are equaled. What do I mean by this? The test guys are trying to fail the part with excitation (within specification limits of course) and the product design guys are trying to prevent the failure by strengthening the area (cause) that was determined to have risk. So try as they might, the test guys will experience a pass not a failure.
Remember we should not learn anything through this test plan as we are trying to verify that the design is acceptable. The design verification should be confirmed or backed up by the test after having first gained confidence in the design through prevention means (standards, FEA, Best practices and Product DNA)
It amazes me as to the usefulness of FMEA across many industries. I am sitting in the lobby of a major defense contractor for the second day of a FMEA risk mitigation event for the Hummer replacement. The new designs are intended to reduce the effectiveness of IED’s or Improvised Explosive Devices. The FMEA is being used to reduce the probability of failure of various systems as they contribute to the potential failure. In other words a systems approach. Some systems have greater responsibility to deliver protection and survivability than others, but all are responsible for some of the functional requirement.
The outputs of FMEA in these examples are extracted from the FMEA and managed through a risk management method embedded within the project management system. This is a great way to get the value from FMEA and then populate the Design Reviews with the actions, responsibility and timing. Follow-up is most likely to occur in these cases. This is a major flaw with using FMEA alone as some folks simply forget to review the actions in a timely way.
Always remember to update the FMEA after the actions have been taken as this assures future considerations of the actions taken in new FMEA development.
Just a note while on the road in New York about prevention and detection controls. Some confusion exists about these two very distinct types of controls.
Prevention controls include things that help you do it right the first time without having to physically test. These include FEA, Modeling technology, Product DNA, and Standards that can be followed. Note that FEA in a iterative design process as opposed to using a the design process. These Type 1 controls affect the Occurrence number and are not to be evaluated for the Detection ranking.
Detection controls or type 2 controls are used to detect design weakness and potential errors. Tests and formal protocols, physical representations of parts, past test results etc… each is rated independantly of each other for each cause and failure mode. the lowest value is used for the detection ranking to be included in the FMEA. More on Process FMEA in my next post
Working in China this week. The FMEA process is just starting to take off in China. The concept of FMEA is always somewhat hard to start with. The design engineers here still struggle with the same issues that design communities everywhere else struggle with. Why do this? I am a good design engineer. I won’t make mistakes.
The message here is the same everywhere: The design engineer should never do their own Design FMEA. Think about it. What will they find? And if they find something, why did they not find it before the FMEA? The design engineer should be doing the FMEA in his or her head as the design process progresses. The Design FMEA is done by a team of people including other design personnel that provide the questions and drive for doing a good job. The design engineer responsible for the design is required for two reasons:
1) They are in the best position to know when the FMEA should be performed: roughly 50% of design completion time or earlier.
2) They need to provide the team with the philosophy of the design direction and choices made.
They must never drive the FMEA as the result will follow the the design engineers lead, and that would diminish the value of FMEA for uncovering possible failure modes, their causes and mechanisms and the actions that will be suggested to make the design better.
Your comments are always welcome!
Just a note while waiting for an airplane to Beijing in Kuala Lumpur. I have been doing alot of work with customers on Special Characteristics lately. One confusing point that requires clarification: what makes a characteristic Special? Is it how important it is? There are several good definitions and it is my opinion several of them work. One such definition is related to control of Variation and if the customer can see value or improvement with this variation minimized. That one is OK, but is this not true of all variation?
I like to suggest it is those characteristics that are at risk of not delivering something important. My example would be:
Is your heart rate important for your survival? Sure it is. Is it special? I think not. Why you might ask? I think it is in control and stable for most of us. Most of us do not get up in the morning and plot a control chart or calculate Cpk on our heart rate. If we are normal (fitting the normal distribution) our heart rate would have a Cpk greater than 2 for sure for what is expected to be normal. However if I were to try to lose 40 pounds by going on a treadmill or walking program, the need to monitor increases due to the increased risk that I just put myself under. A diabetic must monitor and control blood sugar, so that is a special characteristic.
How do I know there may be special characteristics? Design FMEA gives us the best hope of finding them if they exist. When done correctly the DFMEA will give us a clue, but will not be explicit as to where the characteristics are. When should I look for them? Before the design is finished for sure, but also including the manufacturing and assembly peoples input for mitigating the risk by making changes to the Design, Process or both. Error Proofing is our first mitigation obligation, Improved stability and capability the second, and finally special control strategies when risk has not been reduced to an acceptable level. Want to know more? Give us a call at (248) 280-4800 and ask for Michelle.