Basic Dimensions Documented in First Articles 3

[drawoh]

axym,

If I inspect a part, I ought to record the measurements I took. Ideally, I should measure from the datums, but often, I have to work with the tooling I have. Someone, probably not not necessarily me, has to process my measurements and compare them with the GD&T. If there is a basic dimension, there is a feature control frame showing what is allowed. As the inspector, I need to test this, and record the result.

Recently, I inspected some part by sitting them on gauge blocks. When parts were less than the maximum and more than the minimum, I quickly lost interest, and I don't really know what the dimensions were. I found some parts that did not conform, and then I made the effort to determine what the dimensions actually were.

I am not accustomed to FAI's, although I anticipate seeing some soon. I would want to see that actual measurements included in the report.

--
JHG

 

[Vindicit]

I see this exact problem every day. I work as an Engineer of a supplier for major Aerospace companies. The problem seems to stem from the Quality people on both the customer side and supplier side not understanding GD&T enough. From my research, Basic dimensions should be reported but they have no direct tolerance (obviously) so there is nothing to put a direct pass/fail on the individual Basic dimensions. In the past from the likes of GE, Pratt & Whitney and Rolls Royce (all quality people) is that they cant pass a First Article because some dimensions (Basic) dont have tolerances attached to them so their system cant accept them. I've never dealt with any of the Quality systems of any of customer so I dont know if thats true/false. Some customers even go as far as forcing us to impose a made up tolerance on basic dimensions to pass their FAIs which is stupid. I dont know how to solve this issue as its not just one customer doing this.

As for many of the same type of dimensions, such as a ton of hole positions. Usually we are allowed to just give one report line and just put the range that all the inspected dimensions fall between. Ive never had any FAI fail over implied dimensions because every Quality person just looks at a drawing and basically makes a check list of the visual dimensions. If a dimension is not visible on the drawing then they dont do squat for it usually.

But Basic dimensions should be reported on the FAI and with no direct tolerance just as its definition suggests. No Basic dimension should ever fail. If any quality person says for you to do any more than that then tell them to schedule a GD&T course or learn to use their FAI software better.

 

[drawoh]

Vindicit,

Okay. I now have seen some First Article Inspection reports. These are spreadsheets showing the dimensions, the tolerances and the as-measured dimensions. I assume these procedures were developed prior to rigorous understanding of GD&T. The logic is that I measure something. I record the result. I compare the result with the tolerances.

If I inspect a complex outline controlled by a profile, I can easily record pass and fail. It probably is not possible to reduce the test to a couple of recordable numbers. The solution might be prepare an inspection drawing with dimensions with [±][ ]values, with codes referencing the inspection spreadsheet. Obviously, the dimensions would be based on an interpretation of the GD&T on the original drawing.

--
JHG

 

[pylfrm]

If I inspect a complex outline controlled by a profile, I can easily record pass and fail. It probably is not possible to reduce the test to a couple of recordable numbers.

The CMM software I'm (slightly) familiar with will spit out the two extreme values of deviation from true profile, without much more effort than checking a box. If you are using some sort of functional gage, then it may be rather more difficult.

pylfrm

 

[DCarnley]

I'm glad we are having a very good debate on this subject.
I like what AS9102 says using a CMM and how we must program it for FAI purposes.
When automated inspection tooling produces measurement results, those results may be referenced on 9102 Form 3, identified as pass/fail, and attached only when:
[ul]
[li]The characteristic numbers are clearly linked in the attached report.[/li]
[li]The results in the attached reports are clearly traceable to the characteristic numbers.[/li]
[li]The results are directly comparable to the design characteristic.[/li]
[li][/li]
[/ul]NOTE: Coordinate Measurement Machine (CMM) data alone would not be acceptable for a positional tolerance; the results shall show the actual positional value.

So you just can't put CMM Data (X,Y) coordinate you must also show the actual positional value. Also if the feature control frame has MMC or LMC that affect to true position tolerance those should be listed to show why the true position tolerance changed from .010 to .014 for a hole that was .004 larger then MMC.

And to another comment made. And I retort, I am not crippling the power of geometric tolerances. I am all for it if used properly you gain tolerance, control tolerance stack-up and ensure everyone is playing on the same field.... What I am doing is documenting what is on the drawing as design characteristics, yes Basic dimensions are design characteristics, and I give a place for the inspector to put there results. If there are 13 holes to measure I give them 13 places to record it all. Yes a place for each of the 13 holes, 13 for the X's, 13 for the Y, 13 for the diameter and 13 for the position tolerance. I also will index the holes with a lower case letter on the drawing and place that index in the cell so the inspector knows which place to put the result. You might find this overkill but in the interest of why we perform a First Article in the first place is to document the output of the process of making one part of the initial production run. We characterize the part as the benchmark. I can even agree with others that we should add the unspecified 90° surfaces. We look back at the benchmark FAI when something goes awry to see what it measured the first time and compare it to the current state.
Again this is a grand debate and I enjoy reading the thoughts of others on this subject.

Enjoy!

 

[Nescius]

If there are 13 holes to measure I give them 13 places to record it all. Yes a place for each of the 13 holes, 13 for the X's, 13 for the Y, 13 for the diameter and 13 for the position tolerance. I also will index the holes with a lower case letter on the drawing and place that index in the cell so the inspector knows which place to put the result. You might find this overkill...

It may or may not be overkill, if you happen to want the data and believe that data to be valuable. As we discussed above, that data is fuzzy and relies on a number of assumptions.

What I do believe, is that such data is not appropriate on a FAI, because it is not related to whether or not a given part meets its specifications. You are creating a connection with your assumptions, but there cannot be a mathematically certain link because geometric tolerancing transcends the concept of +-/XY dimensioning completely. There is no ability within the language of +-/XY notation to describe the things that a geometric tolerance is describing.

I say that you're crippling the power of geometric tolerances because you're advocating a procedure that is easily shown to be nonsensical now, let alone if we look where part design, manufacturing, and measuring will be in the future. Our ability to design, make, and measure fantastic, organic, free-form, optimized shapes is only growing. XY dimensioning has always been a terrible, mathematically flawed way to describe a 3D part and it's becoming less and less acceptable by the minute.

 

[drawoh]

Nescius,

Assume I have to inspect your part and submit a report and I do not have a CMM. Either I work out a bunch of XY dimensions and tolerances based on your profile, or I create pass/fail inspection tools like a Go/NoGo gauge. I create an inspection drawing. I arbitrarily pick points around your profile and I inspect those points. The FAI is not perfect, but it provides useful information.

--
JHG

 

[Nescius]

drawoh,

As I said to DCarnley, that information may or may not have value. It depends on the individual part and tolerance. I even agree that such information is usually great to have, especially on typical machined parts...which overwhelmingly consist of orthagonal features generated on machines that are very square and repeatable. Even so, that doesn't mean it belongs on a FAI.

If an entity wants to record said information on something they're calling an FAI, that's fine. What really grinds my gears is a customer demanding that I "measure" and record all basic dimensions as part of some PPAP. "Measure" is in quotes because, as I explain above, it is mathematically impossible to distill many things down to an XY pair...a hole "location", for example. "Location" is in quotes for a reason, too...because there's no such thing as a hole "location" in the universe of XY coordinates.

In your profile tolerance example, for XY numbers to have any value, the basic profile must be of constant cross section. Even then, you're assuming the part is uniform in the Z direction, so that your XY at the Z that you chose ends up being representative. This isn't necessarily a terrible assumption, but...

Of course, a profile tolerance can just as easily apply to a complex surface with no constant cross section...with the basic dimensions defined by a model or a mathematical equation. Where does it stop? Thousands of XYZ touch points listed on a FAI? Geometric tolerancing has the power to control an incredibly complex surface with only a single model and a single feature control frame. That is awesome. Being forced to record raw metrology data points, potentially by the thousands, on a FAI...? That is the very definition of crippling the power of geometric tolerancing.

 

[DCarnley]

Nescius,
Maybe you are not understanding my intend of why I record Basic Dimensions on the FAI. True they are not shown as passing or failing for that line item but the basic dimension actuals support the actual positional. They are absolutely necessary to calculate the position. Your argument is that the basic dimensions don't need to be part of the FAI because as long as the actual hole position is within the true position tolerance then no one need know why. This is where we differ. I'm surprised that you think that the basic dimension information is irrelevant on the FAI. I'm not sure you see the FAI as a benchmark for overall process to make the part successfully with data that shows why it was successful. I've reviewed many FAIs from many suppliers. I can tell the difference with suppliers that truly understand why we ask for a first article and other suppliers that just want to get it done from a minimalist approach. It is typically the minimalist suppliers that warrant higher scrutiny by our Supplier Quality Group. I have seen out right lies on a FAI where the inspector has no clue how to calculate position and just puts a passing number. When that same part is measured by the customer they find that the positions don't correlate on a significant scale. From the minimalist FAI (without basic dimensions) the customer cannot determine how the inspector came up with the result. When the part is returned and the customer witnesses a second inspection at the supplier that same inspector shows the lack of knowledge to make the calculation or they find out that out the customer is right and the hole position is not what was reported.

To respond to your gear grinding with customers asking to record the basic actuals for a PPAP, That is the only way to properly analyze the process. Like it or not the basic dimension coordinates are what tell you what the actual position is. Because the theta angle is not recorded I can have a hole with an actual position of .005 and that hole position has an infinite number of location possibilities in a Cartesian circle that the .005 position diameter creates. The PPAP will determine if the process will have repeatability in both the X and Y directions.

I have attached a file using a tool I created in excel to show actual delta position compared to the individual basic dimensions. (deviation from Nominal). Take a look at it and you will see that when plotted interesting things can be ascertained when you can visually see the actual hole positions and the true position tolerances applied.

 

[3DDave]

If I want to know about the inspection I want to know the XYZ and vector the CMM Probe was on at the time of each hit, and the diameter of the probe so I can check that the CMM did its math right. I also expect to see the source code and the compiled code to make sure that what the CMM software developers intended is what's happening, along with any errata sheets for the CPU. Also I need the full temperature distribution of the part and the CMM to make sure they are at exactly 68F.

If that's not on the report how can a person be sure?

It seems to me like you're avoiding qualifying vendor capabilities before letting contracts and using FAIs to inspect quality into the sub-contracting process.

I suppose that's the reason for the disconnect - I would not depend on a supplier with a conflict of interest to supply both the parts and the FAI data without seeing their operation and understanding their process. So why do you accept their FAI at all? Just do receiving inspection and send the failed parts back.

 

[SDETERS]

How are FAI being done on parts that use MBD or 3D PMI, when the basic Dimensions do not appear on a "print" or 3D model.

 

[DCarnley]

SDETER
MDB or 3D PMI requires the Engineer to complete the Model and all tolerancing for all features. Unfortunately this does not always occur for the Model

We do perform inspections on composite material using a GOM Blue Light ATOS 3D Scanner and gave the tolerance based on the Model with a Global Best Fit to a Profile of a surface of .060 (not related to any Datum's). This allowed specialized software to take the cloud of points and manipulate them all together and fit with the tolerance zone of +.030 / -.030 to the model surfaces. This approach does limit any continued analysis as the information is not from any Datum structure.
Don