Inspection and Detection - Kit to production - Need an inspection routine

[Ross_Henderson]

I am getting ready to release some kit to production but need to come up with an inspection routine for this.

Inline monitoring is not a possibility so I need to tag QC onto the end of the process. The machine will output 3000 units an hour and due to the size of the staff the maximum inspection rate will be 1/250. The machine has a problem that causes it to occasionally produce ~10% defective units. The distribution of these defects are random throughout the finished units.

How do I use these figures to calculate the average amount of time to detect this issue or even if we are likely to detect it at all?

Obviously if I can show that employing an extra person to increase our inspection rate will save money overall then I can get this person.

Thanks in advance for you help.

Ross

[Mike S.]

Originally Posted by Ross_Henderson

I am getting ready to release some kit to production but need to come up with an inspection routine for this.

Inline monitoring is not a possibility so I need to tag QC onto the end of the process. The machine will output 3000 units an hour and due to the size of the staff the maximum inspection rate will be 1/250. The machine has a problem that causes it to occasionally produce ~10% defective units. The distribution of these defects are random throughout the finished units.

How do I use these figures to calculate the average amount of time to detect this issue or even if we are likely to detect it at all?

Obviously if I can show that employing an extra person to increase our inspection rate will save money overall then I can get this person.

Thanks in advance for you help.

Ross

Let me see if I understand, please correct me if I am in error:

You only have time to inspect 1 out of every 250 parts the machine produces, or 12 per hour? Why?

You can't adjust the machine to correct it if a defect is found?

The machine sometimes produces 10% defective at random?

Is this right???

[Ross_Henderson]

The machine seems to occasionally damage a piece within the machine, when this damage occurs seems to be random and our equipment engineers are unable to find the exact cause and it seems to happen once or twice a month. When this damage occurs around 10% defects are produced, not every unit is affected and it is random the equipment manufacturers are aware of the issue but have more or less classed it as a feature.

If we increase the inspection frequency we will need to stop the machine in order for the operators to carryout the inspection if we slow the machine down we will create a bottleneck in the process.

Hope this answers your questions.

Ross

[Wes Bucey]

Originally Posted by Ross_Henderson

The machine seems to occasionally damage a piece within the machine, when this damage occurs seems to be random and our equipment engineers are unable to find the exact cause and it seems to happen once or twice a month. When this damage occurs around 10% defects are produced, not every unit is affected and it is random the equipment manufacturers are aware of the issue but have more or less classed it as a feature.

If we increase the inspection frequency we will need to stop the machine in order for the operators to carryout the inspection if we slow the machine down we will create a bottleneck in the process.

Hope this answers your questions.

Ross

I have a sneaking suspicion Mike S. and I are thinking along the same lines.

Here's how I see it, Ross. If you need a temporary fix, consider this:

1. What damage does the machine make when it goes "awry" - same defect or random defects? (This has NOTHING to do with the random distribution throughout the run.)
2. Is this damage instantly visible to the naked eye or is it internal, microsopic, or some other condition which cannot be discerned by vision?
3. Is there a possibility you can screen for JUST the damage on all the parts (100%), leaving your inspection methodology alone? (i.e. Is the balance of the process in control?)

While you do the TEMPORARY fix of detecting defects, you MUST find the root cause of the problem and fix it. Omitting this root cause analysis and subsequent fix by throwing up your hands and letting a machine manufacturer tell you

When this damage occurs around 10% defects are produced, not every unit is affected and it is random the equipment manufacturers are aware of the issue but have more or less classed it as a feature.

is to be a complete traitor to the Quality profession. If the current process cannot be fixed, you need a different process.

Would you give a gun to a soldier and tell him:
"This is a fine gun, but it has a peculiarity which we class as a "feature": Once in a while it develops a quirk and will blow up and kill the person wielding it once out of every ten shots. The anomoly is random. Work with it."

[Mike S.]

I think the answers to Wes' questions are very important.

I might add: You say "When this damage occurs around 10% defects are produced". Does that mean that when the process is "working" there are basically no defects and when it is "not working well" the defect rate is ~ 10% and not 1% or 3%, but always around 10%?

If so, once the machine starts damaging parts "once or twice a month" you should be able to detect this problem with a sampling plan, but sampling only 1 out of 250 will likely not detect it. I have forgotten much of my probability knowledge through lack of use, but I think you'd have to look at about 30 parts with no defects to have a >95% assurance that the machine is not on the blink. Looking at the OC curves of various sampling plans should help you pick a workable sampling plan if you can't 100% test for this defect.

[Sam]

2400 defective parts per shift! I couldn't imagine that your management would even consider running a process with a defect rate of 10%.
Surely you jest.

[dfirka]

Ross,

I agree with the comments made so far, and I think that the problem you described needs more background to be solved without "shots in the dark":

Some additional questions:

1- What is the impact of the effect on the customer or clients ?
2- What is the cost of reparing a defective unit?
3- What is the total cost of allowing a defective unit to reach the customer?

One possibility in the short term is to find an inspection procedure fulfilling two goals:

1- decrease the total cost of poor quality

2 (and very important)- collect data to help in the prevention of the problem, trying to solve the special cause, or trying to find a model to predict the frequency of occurrences of the problem.

Based on your description so far, this could be a very juicy project for any BB.

Pls take this as a very limited answer due to lack of profund knowledge of the problem. Hope this helps.

Daniel

[Wes Bucey]

Originally Posted by Ross_Henderson

The machine has a problem that causes it to occasionally produce ~10% defective units. The distribution of these defects are random throughout the finished units.

Obviously if I can show that employing an extra person to increase our inspection rate will save money overall then I can get this person.

Thanks in advance for you help.

Ross

Say, Ross, if I agree to come help your organization "resolve" its issues, can we go punting on the Thames and carouse in a lot of pubs?