Home > Forums
Topic Title: Multiple 100% visual inspection
Topic Summary: visual inspection
Created On: 01/27/2016 03:39 PM
Linear : Threading
Send to a Friend Send to a Friend
Search Topic Search Topic
Topic Tools Topic Tools
View similar topics View similar topics
View topic in raw text format. Print this topic.
01/27/2016 07:42 PM
Print this message

Author Icon
256003
Jozef Gavel



Hi All,
I have a questions regarding 100? visual inspection.
How to calculate efficiency a pass-fail ration when for same thing are 4 same visual inspection processes.
I know that effieciency of 100% inspection is only 85%, but what efficiencypass-fail ratio will have , 3rd, 4th inspection?
Does exist any formula?

2nd question: Is OK to work 45 minutes and then have short break to have effective visual inspection?

3rd question: Detectability when use FMEA is 4, when we have 2 nd inspection, should detectability be better /3 or same?


Does exist any good manual on the internet?

Thanks.
01/29/2016 07:49 AM
Print this message

Author Icon
QualityJohn
John Niggl



Hi Jozef,

I'm curious about your question...

Are you asking about visual inspection of finished product? Also, how many pieces will be inspected in 100% inspection?

It seems to me that, in most cases, inspection of a sample of product using AQL (acceptable quality limits) would be sufficient.
02/02/2016 12:51 PM
Print this message

Author Icon
Alain0606
Alain Lamusse



Hello Jozef,

Interesting questions as my company also does visual inspection of every product produced (i.e. 100%).

Although this type of inspection is only 85% effective, inspecting the product more times with different people does not necessarily improve the effectiveness. Usually the first inspection will find all the gross defects and some of those that are more difficult to find. Assuming that these are removed from the batch prior to the next inspection, the subsequent inspections will have more difficulty in finding defective product. The degree of difficulty will increase with the 3rd and 4th inspections.

So the scores in your FMEA cannot simply be divided by 3 or 4 depending on the number of inspections. Each subsequent inspection is less effective than the predecessor. I am not aware of a formula so I will be interested to see the results of this question. We have based the score on the overall effectiveness of all inspections being 85%.

Having regular breaks to ensure the effectiveness of the inspection is normal. Our inspections are done under a microscope so the 45 min rule is essential to rest eyes and change the posture. We pay attention to factors affecting concentration, such as ambient noise level, lighting conditions, location to avoid through traffic and other distractions, use of a radio etc

Good luck
02/02/2016 12:51 PM
Print this message

Author Icon
Robert_Simonis
Robert Simonis



The best case is that visual inspection is 85% effective. That is when the inspectors are well rested, with good light and conditions, inspection criteria is clear, and the methods robust. Interesting to note that inspection is more effective when there are about 1% failures. Too many problems and more get through, but less failures and inspectors get bored and miss a failure when there really is one.

The PFMEA rating for detection is a maximum of a 4 for visual inspection and based on your evaluation and data you may decide to have a 4 or less for subsequent inspections. If the first inspection is effective you may a lower occurrence rating for subsequent inspections for a particular failure mode but detection is still 4 or lower. Important to note that many products suffer damage during the inspection process. Second, third, and fourth inspections tend to catch about 10% of the quantity of the previous inspection but different defects than the earlier inspections. For example the occurrence rating for a failure mode of scratches, or surface damage, may be higher as you inspect more.

Ideally each inspection would remove one sigma of variation, so if your process is 3 sigma, three subsequent inspections could achieve a 6 sigma level of the defects caused by the original process; but because inspection has additional failure modes, you could have more defective products reaching the customers than if you had not inspected at all.
02/02/2016 12:51 PM
Print this message

Author Icon
Robert_Simonis
Robert Simonis



If your goal is to certify that your product meets a certain percentage standard, then random sampling can achieve this. If your goal is zero defect, than only 100% will do.

Shingo wrote his first books extolling the value of random sampling but by the end of his writing career he admitted it was a mistake and that true quality can not be achieved using sampling methods. By definition, error proofing, or poka yoke, checks 100%.
02/03/2016 11:27 AM
Print this message

Author Icon
304369
Keith Lodahl



Poka Yoke checks 100% by buildiing the check into the process step. Quality is built in at the source, not inspected in wastefully after the fact. Quality at the source is by far the best way.
02/05/2016 04:35 PM
Print this message

Author Icon
Robert_Simonis
Robert Simonis



I need to correct myself. I do not have the latest rating tables from AIAG, but ten years ago the BEST ranking visual inspection could get for DETECTION was a 7 - that was for double visual inspection. Single visual inspection was no better than an 8. Triple inspection doesn't improve things beyond double.

For those unfamiliar with the AIAG FMEA methodology, the rankings go from 1 to 10 with 1 being the best case and ten being the worst. In the detection category, a one would mean that the failure can not occur - it has been error proofed in the design. A ten would mean you have no way to detect the failure at all. Visual inspection is only slightly better than that.
02/05/2016 04:35 PM
Print this message

Author Icon
Robert_Simonis
Robert Simonis



The best way is to have the failure removed at the product design stage and then it doesn't need to be caught anywhere in production.
02/22/2016 10:50 PM
Print this message

Author Icon
40183
Lonnie Wilson



Hello Jozef,

I presume you are speaking about 100% inspection of finished goods, presumable to avoid shipping bad product to your customer. If that is the case, I have several comments. First don't even consider using AQL methods, unless your customer is interested in gettgin defects. The AQL methods virtually guarantee a certain level of defects. second they really don't have any continuous improvement methodology and third, they are based on a false premise and that is that your process is statistically stable. And unless you are using SPC to evaluate the statistical stability of your process, it will not be stable. Virtually none are until you make them so.
Second, I think you are missing a major point, or you just failed to ask it. That is, what is the purpose of your inspection?? Sorting I presume. That being the case, and you believe that visual inspection done by humans is 85% effective, I would call you very optimistic. and even if it were so, are you just inspecting for one defect?? if not the 85% sorting effectiveness drops off dramatically. since most defect modes are independently created, the sorting effectiveness of say 5 defects would be 0.85 x 0.85 x 0.85 x 0.85 x 0.85 or 44%....Now how do you feel about visual inspection??? I would not be at all surprised if you have some serious problems with your customers. there is a reason that Dr Deming said we should quit trying to inpsect quality into the product, rather build in quality from the start.....(Roberts comment above) because 44% or even 85% sorting effecitiveness will not protect your customer from your deficient processes. Unfortunately we sometimes get involved too late and that option is passed. But what has not passed is to remember that the inspection data has two process purposes (most of it is only used for financial purposes which just measures how bad you were and does it long after the fact.....now that's really helpful is it not?). Those two process purposes are to avoid sending bad product to the customer...AND to use that data to improve the process. Clearly at some point you need great inspection data, but most systems are so broke that you already know your two or three biggest problems and to get really good inspection data would only serve to better quantify how bad your system really is. What you need to do is to aggressively and right now, solve some of those problems and shrink the problem. Sounds to me that you are so busy mopping the floor (double and triple inspection for example) that you cannot turn off the faucet (solve a few problems). Put more attention on fixing the process rather than trying to get good at something you should not be doing at all.

So Jozef my advise to you is to quit worrying about the merits and math of double and triple inspections, quit looking for a manual as they will give you no magic solutions. Your team needs to get busy with basic problem solving and shrinking the size of your problems. If you can then get rid of a couple of your problems, improving your inspection system will be easier. This will not be done with second and third tier re-inspections but will come about by having good operational definitions of the defects along with good training and tools for the inspectors and then you must use that data to drive good problem solving and you then have a fighting chance of succeeding. Virtually everyone who has been in this business for more than a fortnight, has been through this phase on their way to excellence...good luck

Lonnie Wilson
law@qc-ep.com
06/15/2017 01:15 PM
Print this message

Author Icon
342540
Brian Hall



Lonnie,

Your logic is flawed; for using your example if one were to have an infinite number of inspections then the limit of .85 ^ n as n approaches infinity would be 0% effectiveness. While not entirely accurate, the better way to approach it is to the probably of a defect escaping and multiply that to the power of n (0.15 ^ n).

The true problem with multiple visual inspections is the human factor. The feeling that 'if i don't catch it, someone else will' removes the criticality of the inspection from the mind of each inspector.
06/20/2017 05:13 PM
Print this message

Author Icon
Robert_Simonis
Robert Simonis



Regardless if the operator feels it is critical or not; if you are make enough nonconforming parts, some will escape your process and reach the customer (the colloquial way of saying that is "if you throw enough sh*t against the wall, some of it will stick") . Error detection methods, whether automated or manual, can be a way to check your error prevention methods, but by themselves are not an effective method to prevent failure to the customer.
09/18/2019 08:24 AM
Print this message

Author Icon
johnyshaw
johny Petrik



Quality inspection is measures of product quality characteristic to confirm quality
Note: These forums are moderated by the Lean Enterprise Institute. All posts are reviewed within 24-48 hours prior to appearing on the site. Views expressed in these forums do not necessarily represent the views of the Lean Enterprise Institute.