Measurement Resolution of In-Process Automated Inspection System - Stamping House

[Phil P]

Hello Everyone,

I have a dilemma that I hope someone may be able to hep me with. I am currently investigating the introduction of in-process automated inspection of some of our products (we are a stamping house) and am unsure how to verify that the resolution of the proposed system is sufficient. To clarify my intentions I must state that we also intend to perform regular off-line inspection, as not all necessary parameters can be tested in-process. Off-line inspection is performed to the neccesary four decimal places.

I understand that a general rule of thumb is that any measurement device should have an extra degree of resolution to that of the specification, but this will not be practical for us as the specification is to three decimal places and we cannot find any systems that would be accurate to four.

I have spoken to one of the potential manufacturers he has asked me to check the standard deviation of the parameters to be tested. He has provided the resolution and standard deviation of the measurement system. Should I ensure that the standard deviation of each parameter is greater than the system resolution?

Another potential problem I forsee is applying SPC to the process. If the standard deviation of the parameter is smaller than the resolution of the system I believe that we'll get chunky data, which will not be suitable for analysis.

I'm also a bit worried about the system standard deviation, as the quotation states that it is dynamic. I assume that because this has been stated as 'dynamic' (I haven't come accross that before) it opens the door for aproval of the system with massive variation in results.

Any comments much appreciated.

Best regards,

Phil.

 

[Marc]

Anyone knowledgeable in this area?

What's the tolerance?

 

[Wes Bucey]

As I understand "in-process" inspection, the ultimate purpose is to confirm the process is making conforming parts by checking at various stages of production to help avoid continuing work on parts spoiled in an early stage of production. The results from the in-process inspections can be charted to assure the process is in control.

So, from the way the problem is stated, I presume the ONLY question is whether the discrimination of the in-process testing instruments is sufficient for the tolerances of the work product. It appears to be a "given" that the capability of the stamping dies and machinery is sufficient for the product. If it were I, I would go with the in-process inspection at present level of discrimination as an on-going check against "non-conformance" and depend on more sophisticated inspection down the road as basis for control charts.

Often, an additional purpose for in-process inspection is to inspect and chart features that will be hidden or impossible to measure at a later stage in production.

Depending on the feature being measured, the quantity of pieces being measured, and available budgets, in-process inspections can incorporate relatively sophisticated instruments, including machine vision and coordinate measuring machines, but it is doubtful any of those can match the speed of a high speed punch press and would require "samples" (of each process stage) to be kicked out and measured at more leisurely pace. This would take away the "real time" measurement of less sophisticated measuring devices, but would be valuable in creating valid control charts. The simple question is whether the economics of the operation and whether the process, itself, would allow taking pieces out of production and essentially "spoiling" them by preventing them from going through the entire production process.

 

[Phil P]

Hi Wes,

Thanks for your response. To clarify my intentions I want to introduce real-ime inspection on the parameter that causes the greatest amount of adjustment in-process. This adjustment is typically required after material changes (there are regular changes due to the speed of the press, we are also exploring the feasibility of increasing material coil sizes, but that's a seperate issue at the moment) or after the press has warmed-up. We currently apply SPC to the process by collecting samples at the end of each reel of parts.

Really I have one main issue:

1. The resolution of the proposed in-process system is 0.004mm (1 sub-pixel), but the standard deviation of the parameter to be monitored is smaller than 0.004mm. Does this mean that applying SPC to the results would prove ineffective?

A second issue (detailed below) I can address with the potenial supplier once I have the answer to the first, but again if anyone has come accross this before you're input is greatly appreciated!

2. The system in question is a camera system. The quotation states that the standard devaition of the system is nominally 1 sub-pixel (0.004mm), but also that this nominal standard deviation is dynamic. Obviously all standard deviations are dynamic in the sense that they change as more data are collected, but I'm suspicious that this may have been cleverly worded.

All help greatly appreciated!

Best regards,

Phil.

 

[tracy5000]

With the resolution of the measuring equipment the same as your current standard deviation, I would think the resulting standard deviation would be different (I was going to say larger, but might become smaller). You can look at this by taking some of the older data that you have collected and rounding to the nearest .004 and look at the data. I'm assuming that the tolerance of the dimension is at least .015 or you would not be in control. If the tolerance is actually more like .040, than you should be fine with the resolution.

In addition, the standard deviation would be dynamic due application specific issues like environmental conditions (oil from the stamping press or parts potentially getting on the lens and distorting the image), part orientation (distortion from looking at the part on an angle versus straight on), etc. You should be able to question the manufacturer exactly what they mean by dynamic - your salesman may not know, but should be able to get in touch with someone that has more technical knowledge. I they can't give you a good explanation, ask for the test procedure used to calculate the stated resolution.

Tracy

 

[Wes Bucey]

First things first: When in doubt about how ANY manufacturer intends the meaning of a phrase in the advertising or engineering texts, simply ASK the question until you get a satisfactory answer that assures you that you understand what it means and how it will apply to your situation.

I had pretty much guessed the in-process inspection would be vision-based.

Design of Experiments (changing processes) is really only effective once the original process is in control. Do the variations due to material change or speed represent a momentary "blip" or do they continue until an adjustment is made? This is all part of a root cause investigation for either special or common causes of variation. The in-process inspection will let you have real-time tracking of whether the variation is common or special cause. You still have to determine WHAT the exact factor is which causes the variation and eliminate or account for it in your adjustment process. (For example, does the machine heat up and parts expand when first starting up? Perhaps you can install a heating device to maintain a constant temperature. When changing coil stock, are you certain the thickness of the coil stock is uniform throughout the coil? When changing the speed of the press, what happens which causes variation? Is there more vibration? Does the coil stock jump around because of a sympathetic vibration set up [ever see the video of a bridge that started to sway in a high wind, continuously increasing the amplitude, until the bridge literally tore apart?]) Based on your elaboration of the problem, it seems the primary value of the in-process vision inspection would be in tracking the start and stop points of variation, with a view toward reducing or eliminating the causes of variation, not in tracking the effectiveness of your adjustments.

The root causes of your variation may be any or none of the things I suggest. The point is, finding and eliminating the root case of variation and establishing control is necessary before you can even consider running SPC to monitor the effectiveness of the control.

My curiosity is piqued. What product do you stamp with such small tolerances?

 

[Jim Wynne]

The point is, finding and eliminating the root case of variation and establishing control is necessary before you can even consider running SPC to monitor the effectiveness of the control.

Whoa, there. How is "control" established without statistical data? In other words, how do you "establish" control before its "effectiveness" is monitored?

 

[Wes Bucey]

Whoa, there. How is "control" established without statistical data? In other words, how do you "establish" control before its "effectiveness" is monitored?

SPC is for a process in control. Control charts may be valuable for establishing control.

The underlying concept of statistical process control is based on a comparison of what is happening today with what happened previously. We take a snapshot of how the process typically performs or build a model of how we think the process will perform and calculate control limits for the expected measurements of the output of the process. Then we collect data from the process and compare the data to the control limits. The majority of measurements should fall within the control limits. Measurements that fall outside the control limits are examined to see if they belong to the same population as our initial snapshot or model. Stated differently, we use historical data to compute the initial control limits. Then the data are compared against these initial limits. Points that fall outside of the limits are investigated and, perhaps, some will later be discarded. If so, the limits would be recomputed and the process repeated. This is referred to as Phase I. Real-time process monitoring, using the limits from the end of Phase I, is Phase II.

Process Control is the active changing of the process based on the results of process monitoring. Once the process monitoring tools have detected an out-of-control situation, the person responsible for the process makes a change to bring the process back into control.

1. Out-of-control Action Plans (OCAPS) detail the action to be taken once an out-of-control situation is detected. A specific flowchart, that leads the process engineer through the corrective procedure, may be provided for each unique process.
   
2. Advanced Process Control Loops are automated changes to the process that are programmed to correct for the size of the out-of-control measurement.

If the process is out-of-control, the process engineer looks for an assignable cause by following the out-of-control action plan (OCAP) associated with the control chart. Out-of-control refers to rejecting the assumption that the current data are from the same population as the data used to create the initial control chart limits.

For classical Shewhart charts, a set of rules called the Western Electric Rules (WECO Rules) and a set of trend rules often are used to determine out-of-control.

Seems as if guy already knows he has variation, else why is he "adjusting?" Apparently, the variation is not "subtle" or "intermittent," but has definite and identifiable "trigger points" even if we don't know the exact mechanism causing the variation.

The one nagging thing (in the back of my mind) which I hope we can eliminate is that we are not dealing with a "funnel experiment" with gratuitous "adjustments" to the process which exacerbate the situation.

Caveat:
I may be putting the cart before the horse by assuming Phil is looking to verify process capability (checking control limits against specification limits) which cannot be accomplished until the process is in control. There is, after all, normally a difference between control limits and specification limits, or we would all be running a cpk of 1.0.

 

[Jim Wynne]

A few questions: What do you hope to accomplish with the vision system? What makes it potentially better than what you're doing now? What are the attributes that make it attractive?
I understand your concern about the capabilities of the system, but I also understand that manufacturers are loathe to assure capability when they have no control over how the system is used. You are wise to be asking a lot of questions before spending money on what could turn into an expensive white elephant, but you need to be aware that whatever the manufacturer tells you, their results are either wholly hypothetical or based on tests necessarily run under tightly controlled conditions, and those conditions might have been biased towards favorable results. In the fine print of the contract, specific performance claims made in marketing literature are probably disclaimed for purposes of end use, and you will be left with a standard parts-and-labor warranty.

I'm also a bit worried about the system standard deviation, as the quotation states that it is dynamic. I assume that because this has been stated as 'dynamic' (I haven't come accross that before) it opens the door for aproval of the system with massive variation in results.
Phil.

You're correct. "Dynamic" is marketing mumbo-jumbo in this context, and it means "can change radically without notice, depending on any one of an infinite number of variables, or an infinite number of interactions, none of which we can control. Your mileage may vary."

 

[Phil P]

Hi Everyone,

Thank you for the responses.

Wes,

To satisfy your curiosity the parts are used in very samll capacitors, so we manufacture to tolerances as low as 0.05mm.

With the exception of one, all parameters are blanked (so the only source of variation is wear) The other parameter to which I refer is a deflection (the one controlled by the 0.05mm tolerance), and the forming stage of the tooling is designed to allow adjustment to accomodate variations in base material properties. Although it may be theoretically possible to tighten the material specification to reduce the impact of material changes the resultant increase in material cost would far outweigh the negligable value of scrap we produce. It is a single-stage process.

The process is in control, we know this as we have analysed our results for a few years now, for the last year we have come as close to real-time SPC as we can without introducing in-process inspection. From this analysis we know that the defection is the only parameter which changes with these material changes. There is a pattern to these changes, they are also related to temperature (another issue we are currently addressing). But this is not a concern to me at the moment.

My objective in introducing this system is to reduce required level of resources in inspection department, whilst increasing production efficiency through less downtime waiting inspection results.

Tracy5000,

Thanks for your response, my main concern was that the resolution of the system would be insufficient to allow us to perform SPC on parameters with standard deviations lower then the system resolution.

Thank you all for taking the time to help.

Best regards,

Phil.