Visual Inspection Not 100% Effective - How to Improve Visual Inspection

[malasuerte]

Have you found out which defects are escaping to the customer? Establish a pareto of which defects are being missed during the sort. Determine the commonality, if there is one.

Not all defects or nonconformities can be identified using the same process. I would look at your process/es and determine if you have implemented a one size fits all scheme. Could this be your problem?

Also, I am sure you have identified where these defects are occuring(what point of the line)...Rather than doing multiple sorts, are you doing inspections at these points in the line?

A simple tool as well...do you have a database of pictures showing the defects and as per my comment above a corresponsing BKM for finding those defects?

 

[CarlDaniel]

Have you found out which defects are escaping to the customer? Establish a pareto of which defects are being missed during the sort. Determine the commonality, if there is one.

Not all defects or nonconformities can be identified using the same process. I would look at your process/es and determine if you have implemented a one size fits all scheme. Could this be your problem?

Also, I am sure you have identified where these defects are occuring(what point of the line)...Rather than doing multiple sorts, are you doing inspections at these points in the line?

A simple tool as well...do you have a database of pictures showing the defects and as per my comment above a corresponsing BKM for finding those defects?

The large percentage of defects we're having are bent leads & cosmetic defects. I'm not exactly sure what do you mean with a one size fits all scheme but it seems like my problem is the opposite. We have multiple package types of ICs and each has its own criteria, which I think would be too much for an inspector to memorize as part of the semi-annual certification process.
I do have some pictures & drawings of defects on our work instructions but I think it is not enough. So your suggesion of having a database of pictures showing the defects is the way to go...but collecting these samples will require some work.
Maybe I'd look at our certification process which focused on written exams instead of actual assessment of their skills to find defects.

 

[AlanC]

The comment is correct, but having spent man a year at Toyota focusing on this very issue then maybe I can give some ideas
ensure the standard is clear, this sounds easy, but do you really have it well defined. It should be zero defect for first line visual inspection, eg no porosity, no scratches, no burrs. Is the check method defined. If you check every operator do they all do it the same way. For eample, if it was a pen, do they look at the cap first for 2 seconds then rotate it 90 deg to look at the barrel for 3 sec left to right. You need to define the length of time and the exact order of the visual check. QA should audit the visual inspection is followin the inspection sequence and the timings are correct, each person on every shift should be audited. If it is a burr then specify to use a standard pointer. It helps to mark the checked area,bringing the eyes to follow the check. If anything is no good or different to previous then is should no be judged on line but pulled off line to overcheck to separate standard, on line zero porosity, off line 0.5mm in area between x and y.
Consider also the whole picture. The defects you are missing have you mapped them out on a drawing of the product, maybe there's a theme, is the lighting you use set/defined at a certain level eg 100lux. Do you rotate inspectors at least every 2 hours. Does the supervisor ovecheck there work, is there clear feedback when a defect is missed. Other areas to consider, are the problems you miss due to non standard production processes, eg drill break, tool change, in which case this is a different subjec. Anyway I'll stop before a write a book. Hope it helps
alan c

 

[helenqingdao]

in my opinion "zero defect" means "do things first time right", so you'd better check your produce process and try to eliminate defects on line, thus your 100% inspection problems could be solved

just personal opinion and for your reference...

 

[AlanC]

zero defect does men that, from my humble pie experience, internal and external( supplier) defect cause is approx 70% visual, this hasn't changed much over the years, perhaps slight reduction if anything. Most of these causes are non standard, like tool change issues, off loaded stock fed back in at wrong process etc, it depends on the type of defect you are trying to solve

 

[CarlDaniel]

Well put AlanC...very informative. The extent of your controls are well suited for automotive industry.
Ensuring that each inspector is doing it the same way minimizes process variations. So I guess I have to clearly define the best way to do it and make clones out of the inspectors...they should always run on auto pilot right? Auditing their inspection methods is not part of our process so I'll add this also.
Thanks...

 

[Britman - 2012]

The large percentage of defects we're having are bent leads & cosmetic defects. I'm not exactly sure what do you mean with a one size fits all scheme but it seems like my problem is the opposite. We have multiple package types of ICs and each has its own criteria, which I think would be too much for an inspector to memorize as part of the semi-annual certification process.
I do have some pictures & drawings of defects on our work instructions but I think it is not enough. So your suggesion of having a database of pictures showing the defects is the way to go...but collecting these samples will require some work.
Maybe I'd look at our certification process which focused on written exams instead of actual assessment of their skills to find defects.

Morning – I have in the passed been involved in 100% inspection as a manager and a inspector, what I find during the inspection period you being to inspect in the grey areas and as you progress the human brain drifts, this acceptable grey area increases with volume.

An example you start with a component that is only just within the visual / dimensional tolerances, this of cause is acceptable, however as the inspection process moves on a second sample occurs only slightly worse than you passed early, however visually not that much different to the one passed, I have found the brain makes the decision to pass, as the inspection continues this “grey” acceptance standard drifts, until the point the inspector questions him / herself, at this point its to late for many reason to stop the process and retrace earlier passed components – the cause of this in most cases is the believe of failure and fear of management.

As commented at the start I have been in this situation as an inspector and management.

From your information I understand you manufacture IC’s for the electronics industry a couple of questions:

 Do you manufacture to an international standard?
 Do you supply to more than one customer?

If you supply to more than one customers our their comments / defective results similar?

Some corrective plans have been suggested in this thread, best solution is root cause review and achieve product of a known and accept standard.

The agreed standard should be discussed with the customer / s, I would use Physical visual samples if no profession photographs were available, these could be contained in there protective casings – provide your customer / s with near identical samples – this provides a known “base standard”.

If practical I would rotate inspectors, I’ve been there, after long periods of visual inspection on similar product, the brain starts to make its own standards, mainly moving further into the “grey zone”.

For information on Electronic standards review the following site www.ipc.org , they provide a very detail range of publications for the whole industry

 

[malasuerte]

The large percentage of defects we're having are bent leads & cosmetic defects. I'm not exactly sure what do you mean with a one size fits all scheme but it seems like my problem is the opposite.

Sorry, I was making sure you don't have a one-size fits all process in place now.

We have multiple package types of ICs and each has its own criteria, which I think would be too much for an inspector to memorize as part of the semi-annual certification process.
I do have some pictures & drawings of defects on our work instructions but I think it is not enough.

So your suggesion of having a database of pictures showing the defects is the way to go...but collecting these samples will require some work.

The work invested will be worth it. I am in the semiconductor industry. Our database of defects is huge, but it is worth it.

Maybe I'd look at our certification process which focused on written exams instead of actual assessment of their skills to find defects.

Absolutely, One of the most important parts of training is its effectiveness. You need to be able to measure proficiency, not just memory. This is one of my orgs biggest challenges with so many employees. But we find that when proficiency is demostrated we have greater success than tests.

Also, You really do need to understand where these defects are coming from. Inline defect inspection needs to be incorporated. If we had the issue you are talking about, our org would go into a taskforce to find out where and how these defects are coming about. They would not convene until the issue is fixed and preventative measures are put in place.

Also, as said before, conducting both a self assessment and an internal audit of the process could be of benefit.

 

[NAOTB]

I am a Quality Manager at a printing company so we are very heavily dependent on visual inspection.We are undergoing a Lean transition that is approx. 10 mos. old. If you can produce 1 part/ Sheet at a time and inspect as you manufacture you have a better chance at catching defects. If you batch produce and therefore batch inspect in large numbers you increase your chances of missing defects due to boredom and eye fatigue.It's very hard to change our production philosophy but we have proven that it works.

 

[CarlDaniel]

After reading numerous threads on visual inspection I found out that there really is no single formula one could use to improve the inspection effectiveness. I guess when human factor is involve there are too many variables to consider.
I know you guys have a lot of ideas in store but here are the common denominators:
Training - ensuring that inspectors know what they are looking for and that they do it the same way all the time.
Visual aids - a database of different types of defects and master samples (which also needs to be checked periodically).
Auditing - ensuring that the inspectors are still doing what they're trained to do.
Measuring effectiveness - many different ways to do it...I'm planning on using the (Cohen's) kappa found on the MSA manual to evaluate our inspectors.