Guidelines for in-process inspection of PCB repairs:Changing from end-item inspection

[The Fixer]

Let me begin saying I am impressed with the expertise exibited in this forum. I am NOT an expert so please utilize KISS principles if you respond.
My company has 21 branches engaged in the repair of printed circuit boards in telecommunications equipment. Long before I got into the quality organization, someone decided that end-item inspection of workmanship, packaging, cosmetics, and functionality using MIL-STD-105 was the way to go.
I think this whole process needs to go ! End-item inspections:
a) attempt to "inspect" quality into the system
b) cause a QC bottleneck
c) waste time with lot resubmittals via rework and reinspection
d) allows technicians to submit substandard work hoping that their units will not be selected as part of the sample.
e) etc., etc., etc.

I am looking for guidelines to perform work-in-progress inspections so quality is built into the system, not retroactively inspected into it. However, 99% of the information in this forum deals with high volume manufactured product. In my business, all the six-sigma, cpK, characteristic curve stuff was performed on the product when it was originally manufactured. We are basically doing rework when the unit fails in the field.
I am having difficuly determining among other things, (with respect to in-process) how to define a "lot" and "sample". Some processes involved are checking the unit in, following a documented repair process for each unit type, affecting repairs in accordance with IPC-610 standards, proper packaging and labeling, modifications, and shipping.It's seems as if I need an almost perpetual internal audit program.

Allowing for my lack of SPC insight, understand that I have the customer's best interests driving this request.

I'm not looking for a system so riddled with data analysis that it negates the time savings I'm trying to gain by eliminating the "extra" process of end-item inspections. Any arrows pointing in the right direction will be greaty appreciated.

[Marc]

I am looking for guidelines to perform work-in-progress inspections so quality is built into the system, not retroactively inspected into it.

I suspect you are going to have a bit of a problem defining 'work in progress' inspections. Typically in repair you have a variety of possible problems. If you have 21 branches you probably repair a somewhat wide variety of products. I suspect you are hiring folks who are 'qualified' to troubleshoot. Obviously I have no idea what your processes are. Do the techs have automated test equipment? Do they have OEM prints?

You talk about Mil-Std-105 which is a sampling 'procedure'. I'm not sure exactly how you would apply that to your end item inspection. Are you saying you take a sample and examine "...inspection of workmanship, packaging, cosmetics, and functionality..."?

I suggest you determine what checks you currently do at the 'end' of your process and determine where each event occurs. I'm going to assume that a functionality test/check would be made by the technician doing the repair - not at the end of the 'process'.

What I am saying is provide a 'router' with appropriate checks/tests at each stage. You say: "...d) allows technicians to submit substandard work hoping that their units will not be selected as part of the sample..." This is a factor of your company's tenor. If you allow this for any reason you're out of control. I'm not saying this is not typical - many companies are. The question is why? If a tech submits sub-standard work, what happens? Does this affect the tech? Is the tech called to explain what happened? Is there a mechanism to track what techs consistently submit sub-standard work?

What are your current 'end of the line' inspections/tests telling you? What are you finding and how are you addressing the issues?

I am having difficuly determining among other things, (with respect to in-process) how to define a "lot" and "sample".

I don't know that you have 'lots' in your business in the traditional sense. As far as a sample goes, you have to look at your process. Give me an example of what you are currently sampling for.

It's seems as if I need an almost perpetual internal audit program.

You internal audit program should be verifying (by sample) that employees are doing their job in the manner you specify in your documentation and systems in general. I'm not sure what would drive you to a 'perpetual' audit program. Internal audit programs are, by definition, 'perpetual' (you do them every year) but not every day.

Allowing for my lack of SPC insight, understand that I have the customer's best interests driving this request.

What do you currently track statistically?

[The Fixer]

This is how it works: (broad brush)

Defective units are received and a Repair Order is generated.

The units are filtered into the "tech room" by dept based on unit type.

The units are diagnosed, repaired, and modified as necessary by technicians using both company designed and OEM automated test platforms.

Repaired units then go to labeling.

NOW, the units are placed on a "Awaiting QC" cart and inspected in lots by single or combined department number.

The sample size is determined by MIL-STD-105 for lots usually ranging from 30 - 150 units.

The QC inspectors, AT THE END OF THE "LINE" now inspect for packaging, cosmetics, workmanship, and functionality. (The samples have to be taken back to the test platform and run by QC inspectors)

Then the units are taken to the warehouse area for return to stock and eventual shipping.

The QC function stuck at the end of the process adds additional turn around time. Also, the technicians have to move to another function while QC has their bench for QC testing.

This whole process is reactive and time wasting. You can see why I want to change it.

By the way, we have been registered to ISO9001 since 1995 and have had very successful registration and surviellance audits. I'm not trying to imply that there is no quality conciousness here. I just think that you can only improve the quality of work so much with this end-item approach.

I want to have the QC inspectors on the floor auditing all the functions in real time, not after the fact. That's what I meant by "perpetual internal audits".

Yes, I doubt that "lots" would be relevant and therefore MIL-STD-105 would no longer apply. I have gone out to the floor and experimented with this and found a host of issues that never came up during the end-item inspections. I found I was fixing the cause, not the effect. This is exactly the goal of what I termed an "In-Process" inspection technique.

I posted to this forum to see if there were any guidelines with which to govern a non manufacturing inspection scheme like this.

If not, maybe I'll just go ahead and invent it !


Thanks,
Glenn

[Sam]

My opinion follows each step of "The Fixer's" process:

1- Defective units are received and a Repair Order is generated.
a- Tech verifies and signs off as complete

2- The units are filtered into the "tech room" by dept based on unit type.
a- Estasblish a designated hold area. Release work based on FIFO.

3- The units are diagnosed, repaired, and modified as necessary by technicians using both company designed and OEM automated test platforms.
a- Develop checklists that the Tech can sign off to indicate verification/validation of repair requirements.

4- Repaired units then go to labeling.
a- Tech sign-off

5- NOW, the units are placed on a "Awaiting QC" cart and inspected in lots by single or combined department number.
a- Eliminate this step.

The sample size is determined by MIL-STD-105 for lots usually ranging from 30 - 150 units.
a- Archive the MIL-Std-105

6- The QC inspectors, AT THE END OF THE "LINE" now inspect for packaging, cosmetics, workmanship, and functionality. (The samples have to be taken back to the test platform and run by QC inspectors)
a- Eliminate this step

7- Then the units are taken to the warehouse area for return to stock and eventual shipping.

Develop a checklist for your inspector/auditor to use for randomly auditing the process,i.e, two audits in the AM, two audits in the PM.
In-process is completed and documented by the Tech.

Hope this helps a little.

[The Fixer]

Sam,
Guess I should have provided a little more info. The Repair Order goes through the entire repair process from receiving to shipping and contains signature blocks for the Receiver, Technician, QC, Labeler, and Shipper. So everyone already has accountability along the way. Each equipment "handler" is responsible for certain entries on the RO Card. This isn't difficult to inspect. However, we find that the vast majority of errors are in the information that is entered on the RO Card. (Wrong unit ID, Wrong repair code, Wrong warranty status, etc.)
By the time the unit reaches QC, the errors have already been committed. Most get fixed as the unit moves through the process. But some get through the whole process. Those are the ones that QC finds. That's where I think the inspection of work-in-process would be valuable. We would have a better chance of correcting the root cause so NO-ONE will have to fix someone elses mistake.

The biggest problem I'm having is structuring the auditing of the repair process. A lot of the technicians' job is cognitive. We have no Quality System Procedures which define the techs diagnostic thought processes. So, if I am auditing the testing process of an "ABC" unit, there will come a time at which the tech will utilize cognitive skills to diagnose the problem. So what does the QC inspector do at this point ? Go to another bench and come back later ? It could take the tech a couple minutes or an hour to figure out what's wrong with the unit.

The other processes (receiving, labeling, etc) are not difficult to monitor because they follow an established routine. But trying to observe the repair process is tough.

But someway, our process has to ensure reliability before this equipment goes to the customer. I just don't see a way to get around re-testing by QC at the end of the process.

[Marc]

When you analyze what 'gets past' your technicians, what do you find? That is what would be my initial guide.

[The Fixer]

Marc,
Your last response evoked an interesting theory which I haven't seen presented as a part of common statistical analysis techniques.
First, I have created a national database via which I can monitor QC data in any dept, branch, region, or nationally. I use primarily a PARETO analysis or a parameter search to examine specific QC stats.
Now, the theory. Looking at the data over the past year or so, I've concluded that our Quality System has reached what I will term our PME - Point of Maximum Effectivity.
Given the "society/environment" in which a quality sampling technique exists and the specific type of technique used, is there a point (practical not theoritical)at which there can be no more improvement ? We seem "locked in" to relatively the same types and quantities of errors regardless of continuous improvement initiatives.
Considering your last response, I'm convinced that although I had not looked at it from this perspective before, my perception of the need for a change in inspection methods is due to the fact that management involvement, internal audits, corrective actions, etc. have made "this" system as effective as it's ever going to get.
Specifically addressing your response, the PARETO charts can certainly go a long way in determining what to "look for" when performing "work-in-progress" inspections. Unfortunately, due to the number of functional failures showing up in the charts, I don't think I'll be able to eliminate the need for the "invasive" process of having QC re-test finished product. It's just too high a risk to the customer at this point. I'll have to just wait and see if new inspection techniques eventually reduce functional failures by embedding a higher quality mindset into the repair process.

Glenn

[Marc]

My overview of quality assurance and all that jazz is that of an evolutionary process. That combined with continually hypothesizing and testing. The rut is the problem for many. It becomes "..This has worked for the last x years..." or "...we tried TQM and it failed..." What I line about implementations, and what makes me shine, is during an implementation (typically) the company is forced to take a hard look at some aspects of their processes. It's not me - I shine because some people 'open their eyes' and it *appears* to be me...

You know, Glenn, your attitude will achieve as much as (if not more than) any tool out there. Question, hypothesize and test.

What this means to me is new ideas. New thoughts. New blood. Look at San Jose - they want kids who are asking questions. Looking. Feeling.

Management says "...implement TQM..." What does that tell one? TQM is a number of tools.

The US repubilcan party fights statistical techniques in the 10 year census. What does that tell us?

If statistics 'grabs' you, they can tell you a lot. But - don't let that keep you from looking everywhere for answers - or from looking for alturnate tools. Use stats to guide your questions and hypotheses. Then use stats to evaluate your test results.

[This message has been edited by Marc Smith (edited 29 February 2000).]