Corrective action
Six Sigma – iSixSigma › Forums › Old Forums › General › Corrective action
- This topic has 20 replies, 11 voices, and was last updated 18 years, 7 months ago by
Ben R..
-
AuthorPosts
-
December 1, 2003 at 2:17 pm #33995
AndrewsParticipant@AndrewsInclude @Andrews in your post and this person will
be notified via email.On an average,25 batches per month are rejected in our plant due to quality problems out of around 400 batches run per month.Our corrective action procedure takes up only top 3 problems that contribute maximum to scrap. Corrective actions are discussed and finalised in the corrective action team meetings held once in a week.Any suggestions for improvement in our procedure???Andy
0December 1, 2003 at 2:40 pm #93071
Dr. TedParticipant@Dr.-TedInclude @Dr.-Ted in your post and this person will
be notified via email.Here is the issue – All 25 batches affect the bottom line of the enterprise. I have faced this in the great world of software development. In that situation, it was only the “top ten” (out of about 30) that were discussed and fixed. However, instead of talking about “batches”, I talked about money. In my situation, the remaining N defects generatged over $900K in customer downtime, manual fixes, call center reports, developer time, etc. to fix. When shown to executive leadership, the remaining defects received a lot more attention. But….in the world of limited resources, not all defects may receive the necessary attention. In my situation, the “work around” became the official resolution for many of the defects.
Reduce your defects to dollars, and you will have a better chance at resolving those batches beyond the top three. I am ASSuming that none of these batches affect “compliance” of “safety” issues.
Dr. Ted
SSMBB
0December 1, 2003 at 3:08 pm #93072
Nitin SahniParticipant@Nitin-SahniInclude @Nitin-Sahni in your post and this person will
be notified via email.Andy
If you are Eliminating top three reasons every month then by now you must be running virtually a scrap free and defect free process……..
How come you are constantly rejecting 25 batches / month.
Your Corrective Action seems to be more of a Correction
0December 1, 2003 at 3:19 pm #93074
AndrewsParticipant@AndrewsInclude @Andrews in your post and this person will
be notified via email.In some cases , the problem is solved but in some cases the problem repeats.We are dealing with almost 100 parts / month having atleast 10 parameters each that have to be met (approx. 1000 opportunities for problems).
0December 1, 2003 at 9:33 pm #93093Andrews,
It would be a good idea to let the readers know what you have for the following:
1 – The AQL levels for the defects (10 defects) you inspect.
2 – The sampling plan you follow.
3 – The 25 batches which you reject (on an average) is rejected for what type of defects – attribute type or variable type.
4 – Have you done any type of measurement system analysis? What type of gauging is being used?
5 – If the 25 batches are not contributing to the value of the scrapped product (that you are saying you focus on to work on the corrective actions) then you are indirectly not correcting anything at all as you never get to those for corrective actions as Dr. Ted has mentioned. Nitin also pointed out that you should have been done with by now with the resolution.
6 – Are these 25 batches rejected the same products, similar products, what?
In order to be effective, you need to look at occurrances of these failures (are these 25 batches rejected for the same quality defect – have you pareto charted these?), look at prioritizing in $$ (as Dr. ted suggested) and then look at your equipment for variablity for solution.
PB0December 2, 2003 at 1:04 am #93102In case of repeated defects, have you tried looking for patterns of defectives coming out of the same machine/produced by the same operator/in the same shift/same source of raw material etc.?
If you can brainstorm with different guys and make separate Cause Effect diagrams for each of the most frequent defects, you can then zoom in on the most likley cause for each defect, which can then be actioned upon.
Talk to the low man on the totem pole. He knows more than you assume about what causes the defect. Dont forget to mention his name to top mgmt once you have solved the problem!0December 2, 2003 at 4:03 am #93108
Mike CarnellParticipant@Mike-CarnellInclude @Mike-Carnell in your post and this person will
be notified via email.Mu,
You started out making a lot of sense in your first paragraph. The next 2 paragraphs are the same soft opinion driven nonsense that damaged TQM.
If you have data why do I need “brainstorming” what some “guys” think so I can stick them in a C&E diagram. Let’s just take the data out of the deal and do some multivoting next. You go where the data leads.
The other thing about asking the operator. To frequently their knowledge is a function of the same poor measurement system as everyone else. If you think this is an epiphany it is old news.
Now read the post. You didn’t answer the question – this is just repetition of some tired old dogma.0December 2, 2003 at 4:10 am #93110
Mike CarnellParticipant@Mike-CarnellInclude @Mike-Carnell in your post and this person will
be notified via email.Andrews,
There is a pretty basic tool called a Paytner Chart that we used to use to track CA. It ties actions to results. It was part of the Ford TOPS training. There was a post about a year ago where some people found an internet reference for me but I don’t have the link any longer.
Track what you are doing and if it is working leave it alone. If it isn’t working do something else.
That was pretty pretty profound – what can I say it is still before midnight.
Good luck.0December 2, 2003 at 4:33 am #93111Hey Mike,
I thought that the new rules for analysis is It doesn’t make it wrong it is just different. That soft opinion driven nonsense that damaged TQM maybe how they do things and even if it might lead to the wrong conclusions, its something. Kind of like your two schools of thought on SPC limit calculations.
No need for standards, right?
Statman0December 2, 2003 at 4:40 am #93113
Mike CarnellParticipant@Mike-CarnellInclude @Mike-Carnell in your post and this person will
be notified via email.Statman,
The difference is that the two ways of handling SPC both work.
The wandering around building, campfires, and singing Kumbaia naked around a campfire has a track record of no results. It is opinion based and has a poor track record.
0December 2, 2003 at 4:44 am #93114No, the correct method always works the other works sometimes…in both situationsStatman
0December 2, 2003 at 5:03 am #93115
Mike CarnellParticipant@Mike-CarnellInclude @Mike-Carnell in your post and this person will
be notified via email.Statman,
Sticking with the data will get you there.
Two different methods of handling SPC gets you there using data and both work. Both stay with the data.
Believing that you can have a discussion with an operator and you have an instant solution is naive particularly when you are pushing a process beyond 3 sigma.0December 2, 2003 at 5:09 am #93116Yes, and even when you “stick with data” there needs to be standards on how to draw conclusions from the data. Otherwise the data is no better than opinion. I have seen as many wrong decisions made with data as I have with opinion. Look at the data used to predict the Titans victory tonight.
Statman0December 2, 2003 at 5:12 am #93117
Mike CarnellParticipant@Mike-CarnellInclude @Mike-Carnell in your post and this person will
be notified via email.Statman,
I am not sure what the analysis was on the Titans but what passes for anlysis on most of these sports shows is flawed at best.0December 2, 2003 at 5:14 am #93118True!
But they are so sure of their opinions (loosely based on data).
Regards,
Statman0December 3, 2003 at 2:20 am #93144Hey, there is nothing wrong with dancing naked singing Kum By Ya – just not much of a way to operate or improve a process.
0December 3, 2003 at 1:30 pm #93162The link below maybe helpful. It includes Paynter Charts in the handbook.
http://www.philipslogic.com/quality/handbook/106-138.pdf0December 3, 2003 at 2:34 pm #93165
Mike CarnellParticipant@Mike-CarnellInclude @Mike-Carnell in your post and this person will
be notified via email.Stan,
Dancing naked is certainly in the finest of Texas traditions as Mathew McConaughey will attest to. Like you said – it is marginal as an improvement methodology.
It does seem to sell well to people from North of the Mason Dixon as a way to re-establish their manhood. Selling snake oil. Does that sound like someone we all know?
Regards,
Mike0December 3, 2003 at 2:38 pm #93166
Mike CarnellParticipant@Mike-CarnellInclude @Mike-Carnell in your post and this person will
be notified via email.Lee,
Thank you for the link.
Regards,
Mike0December 4, 2003 at 6:55 am #93197Andrews,
Collect the historical data and use pareto chart to know which parts need to be prioritized. If you will not do this you will be “boiling the ocean” – your problem will not be solved.
Solve first the part with the highest frequency of contribution to the defective 25 batch product ( I would assume that 1 product has 100 part and with 10 parameter each ).
Focus to that part and do MSA for each parameter of that particular part to make sure that the rejection is not due to measurement system. Then, collect data and compare the following:
a) operator to operator
b) work station to work station
c) shift to shift
d) day to day
e) machine to machine ( if any )
f) etc… that may cause the variation.
Use 2 sample t or anova for continuous data collected and 2 proportion or chi square for attribute data.
When you see the significant source(s) of variance, focus on that source, this is the time you would have the C & E brainstorming but should also validate if those possible root cause which are subjectively voted to be the root causes are significat by comparing again. collect data, data and data.
By doing these steps, you are cutting you problem by 50% all the time until you reach the most possible root cause. then do DOE if applicable and make sure you put a control on your improvement to ensure the problem on that particular part number will not reoccur. use poka-yoke, automation if possible.
then, proceed to the next part or make many six sigma teams to take care of the other parts.
hope this would help,
AbetF
0December 4, 2003 at 7:04 pm #93219Andy,
It may be that you are not defining the true root cause (or causes) in those instances that recur. This is not uncommon and usually caused by a rush to judgment. Sometimes these things just take time.
Your procedure is probably no different than most. At issue may be your company’s skill in applying the tools available for root cause determination and corrective action. Search the web and you will find a half dozen or so different approaches. You can use these to refine your own approach.
Good luck — Ben R.0 -
AuthorPosts
The forum ‘General’ is closed to new topics and replies.