My issues with PCDmis Ver 4.0 & DataSuite 3.33 (please read B&S!)

[Jeff Hunt]

I'm starting to put together a list of issues I am having with the latest software. If any of you have experianced any of these or know workarounds, please let me know. I am planning on forwarding these issues to B&S & Willcox once I get them all together.


-Dataview V3.22

-When forcing directions with using the "variable.Z!" type tag in the variable sets, dataview will not show any variable with a forced direction tag when you want to just show only points that are out of CP or CPK.

-Edge points automotive direction letters are commonly shown in the wrong axis. This seems to be because the deviation on the sample hit of the edge point is greater then that of the edge point itself.

-There is still no way to set Min & Max Subgroups in Dataview. Whenever you have a sample group that is not a multiple of 5, your CP, CPK, PP, & PPK values will all be incorrect because you can not set this. Datapage has the option to set this.

-When a run chart is reported, if there is a hole reported on in 2 seperate axis, the report will show them next to each other, thereby shrinking the graphic of the part. It would be nice to have an option to place them one above the other instead of next to each other.

-The preview of the report looks nothing like what ends up being printed.

-It would be nice to have the option of reporting the variables in alphabetical order instead of randomly.

-Datapage V3.33

-Datapage does not seem to recognize a forced direction tag in a variable set.

-Datapage Editor V3.33

-Every dialogue box that you open looks & acts totaly different then the rest. None of these boxes remember what you last picked (like the datapage & dataview boxes do). The whole interface seems very clumsy & could use alot of refining.

-PCDmis V4.0 release

-After executing a program or measurement, sometimes the report window will open over the graphics window for no apparant reason.

-Square slots do not pick off of the Cad model correctly.

-Sometimes during the execution of a part program, the first tip change is not done. This seems to happen once in a while for no reason at all.

-The DataFieldEdit wizard starts very slow when it is opened. It almost seems not to be working as it is loading up.

-The toolbar layout still gets messed up when PCDMIS is minimized.

[SABarber]

Should it be "variable.Z!"?

[Jeff Hunt]

Should it be "variable.Z!"?

Fixed. Thank you!

[Jeff Hunt]

Heres a new one

-DataView

-Some edge points that have a forced direction are reported with the wrong automotive deviation letter. For example, a edge point that is long & should be shown as "D" instead is reported as "U". Attached is a report showing what I am talking about. Point e4738b017.t is a perfect example. The deviation is positive, which means the trim is long, which in turn means the trim is down, but yet the runchart shows it as being up.

[Chuck Burbank]

First two items...

The main problem is that Dataview has no idea what the surface normal vector is. It makes a best guess by subtracting the nominal XYZ from the actual XYZ. This works fine for surface points, but can cause trouble with edge points, particularly when the flush error much outweighs the gap error.

One way of dealing with this is to force the direction of reporting with an ! in a viariable set. Seems as though there are some issues there.

Another way is to tell DataPage what the IJK of the point is nominally. One way would be to report I, J and K axes into DataPage, but there is a simpler way. You can create a "directory" file. This file has a few uses. For one you can specify pointing coordinates of a feature, If you were reporting surface and/or edge points and did not want to report the X, Y and Z values to DataPage just becuase DataView needs them, you can defined them in a file that will be used at the time DataView creates a report. You can also define the vector that will be used to decide which direction to declare. This should be the normal vector or edge vector. The format of this file when you are defining the vector should be.

FEAT X Y Z I J K

You only need to define the features that you are having trouble withi.

See the description of this function in the DataView help file.

Item 3:

Does that mean you want to use a subgroup size of 1? DataPage does not include partial subgroups, nor does DataView. If you want every point included regardless of the number selected, you want subgroup size 1.

Item 4:

The placement of every thing is calculated based on the size of the fonts defined in the configure system options and the parameters in the configure output options (dense mode, Min Chart Width). if you play with the character sizes and make the Min Chart Width larger, you will probably find a setting that will work for you.

Item 5:

DataView does not have a print preview option. The screen output is different than the printer output due to large difference in pixel size. DataView is not a WYSIWYG. I have a pdf print driver installed (cutepdf is a good free one). I print to a pdf file to see what it looks like.

Item 6:

It doesn't report them randomly, it goes through an algorithm to keep the leader lines from crossing. When a large number of variables are reported, it looks pretty crappy when all the leader lines are crossing each other, it is very confusing. You can force what features go on a page, but not their order on the page. You would do this by defining #VIEW or #ISO in a variable set. See help file for that.


Regarding DataPage and forced direction... DataPage does not give any feedback regarding direction of deviation.

[Jeff Hunt]

[QUOTE=Chuck Burbank]...

Item 3:

Does that mean you want to use a subgroup size of 1? DataPage does not include partial subgroups, nor does DataView. If you want every point included regardless of the number selected, you want subgroup size 1.

[QUOTE]

Item 3: Say you did a 6 sample study. To get the correct values cor the stats, I would use in DATAPAGE a SUBGROUP size of 1, and a MIN SUBGROUP of 6, and a MAX SUBGROUP of 6. All this is input in the OPTIONS - SPC CALCULATIONS - SUBGROUP PARAMETERS of Datapage. Now you print out a whisker chart that shows the CP & CPK values for the 6 pc study.

Now, go in to dataview. For a 6 pc study, what would you set the subgroup size to? If you set it to 1, the CP & CPK values are different then your datapage values. This is because you dont have th ability to set the min & max subgroup size.

Mabey Im missing something, but the way I was tought to set the SPC calculations was to take the number of parts measured and make the subgroup size times the min or max subgroup equal the number of parts. here are some examples:

30 pc study
sub size = 6
min sub size = 5
max sub size = 5

10 pc study
sub size = 2
min sub size = 5
max sub size = 5

6 pc study
sub size = 1
min sub size = 6
max sub size = 6

[Jeff Hunt]

Regarding DataPage and forced direction... DataPage does not give any feedback regarding direction of deviation.

Don't you think that it should though, because it goes hand in hand with dataview?

[Chuck Burbank]

Item 3: Say you did a 6 sample study. To get the correct values cor the stats, I would use in DATAPAGE a SUBGROUP size of 1, and a MIN SUBGROUP of 6, and a MAX SUBGROUP of 6. All this is input in the OPTIONS - SPC CALCULATIONS - SUBGROUP PARAMETERS of Datapage. Now you print out a whisker chart that shows the CP & CPK values for the 6 pc study.

Now, go in to dataview. For a 6 pc study, what would you set the subgroup size to? If you set it to 1, the CP & CPK values are different then your datapage values. This is because you dont have th ability to set the min & max subgroup size.

Cp and Cpk calculations are based on the subgroup size, the spc chart selected (for subsize > 1, XBar&R, XTilde&R or XBar&S, for subsize = 1, RX, XMR(2) or XMR(3)) and the number of pieces of data anayzed. And of course which pieces of data. If you have subgroup size 1 in DataPage and are using the RX chart, and subgroup size 1 in DataView and are using the XMR(3) chart then you will get different values for the calculations. This is because Cp and Cpk use estimates of the standard deviation that are derived from the current chart selection and subrouping settings. The values Pp and Ppk should agree irregardless of the chart selected as it uses the population estimate of the standard deviation in the calculations. There is a discourse in the help file regarding the calculatoin of capability indices.


Mabey Im missing something, but the way I was tought to set the SPC calculations was to take the number of parts measured and make the subgroup size times the min or max subgroup equal the number of parts. here are some examples:

30 pc study
sub size = 6
min sub size = 5
max sub size = 5

10 pc study
sub size = 2
min sub size = 5
max sub size = 5

6 pc study
sub size = 1
min sub size = 6
max sub size = 6


All of these could be accomplished by setting the subgroup size to 1, the minimum number of subgroups to 3 and the maximum number of subgroups to 2000 or some large number, then just select the transaction dates you want to analyze.

I think most people use DataPage and DataView like that.

[Chuck Burbank]

Don't you think that it should though, because it goes hand in hand with dataview?

As you can see, it's more complicated to figure out what is the direction label that is called out for the deviation. It requires some knowledge of the geometry of the part. As far as spatial relationships go, DataPage has no knowledge. The target plots are the only case where DataPage applies any sense of geometry. I don't think that will happen in DataPage, but it's a good idea for the next generation.

[Jeff Hunt]

As you can see, it's more complicated to figure out what is the direction label that is called out for the deviation. It requires some knowledge of the geometry of the part. As far as spatial relationships go, DataPage has no knowledge. The target plots are the only case where DataPage applies any sense of geometry. I don't think that will happen in DataPage, but it's a good idea for the next generation.

Thanks for the explination Chuck. I guess I was taught wrong to begin with. What wxactly does the subgroup, min subgroup, & max subgroup signify? I have a feeling that I should always have it set to 1 at this point.

I check automotive sheet metal stampings such as doors. I do this in sets of 5, 6, 25, & 30. All depending at where we are as far as selling the part off. My customers main concern is PP, PPK, CP, & CPK. What would you suggest my settings be for this?

[Chuck Burbank]

Usually the subgrouping reflects your sampling. Suppose you pull 3 parts that were made at about the same time and measure them, and then come back later and pull 3 more parts and measure them and come back later and grab 3 more... In that case you would use 3 for your subgroup size. It is hoped that all the data within a subgroup reflect the same sate of the process, so that within subgroup variation is the variation of the process itself. If you don't have any particular sampling strategy like that, you should probably set the subgroup size to 1.

The minimum number of subgroups is just a way of telling datapage to not even try to analyze the data until that much data is in the system. The smallest values you can have are subgroup size 1 and minimum number of subgroups 3. So you can calculate stat values with 3 pieces of data. Of course, you'd want to do that.

The maximum number of subgroups is a way of limiting the number of pieces of data. If you have sugroup size of 1 and max subgroups 30 then datapage will not look at more than 30 pieces of data. if you selected 33, it would analyze data values 4 through 33, omitting the first 3. This value was more useful in the original datapage that was created for the old AVAIL and MM4 systems. Those versions of datapage did not have the "select last N" transactions button, they were a little more difficult to pick the data.

Regarding reccomendations... It sounds like you probably want to have your subgroup size set to 1. The min and max subgroups really aren't that important, I'd set min to 3 and max to 1000 or something like that. When you perform an analysis, just pick the data you want from the transaction date list box. Min and max sugroups only play a role in controlling what data is selected, it doesn't play any role in the analysis of the data after selection.

Say you are looking at a 30 piece study. No matter what your subgroup size is, Pp and Ppk should be the same as they depend only on the sample estimate of the standard deviation, not on the estimates made from the control charts. Cp and Cpk do depend on the estimate of the standard deviation made from the control chart, so the subgroup size does make a difference.

Also, the selected control chart type will change Cp and Cpk. It shouldn't be a large change, but it will be a change. If you use subgroup size 1, then the three choices for chart tyhpe of RX, XMR(2) and XMR(3). Which of these you use is up to you, or maybe up to your customer. I believe that Ford prefers using XMR(3). I am not sure about other companies.

Finally, the customer may want you to use a certain subgroup size to agree with the way that they do their SPC. You could ask them what subgroup size they prefer, and which chart type. If they are using a subgroup size larger than one, then the choices are XBAR&R, XBAR&S and XTILDE&R. I think just about everyone uses XBAR&R for subgroups larger than 1.

[Chuck Burbank]

where I just said...

"Of course, you'd want to do that"

I meant to say...

"Of course you'd want to use more data than that".

I think my other typos are easier to understand. :-)

[Randy L.]

[QUOTE=Chuck Burbank]
Cp and Cpk calculations are based on the subgroup size, the spc chart selected (for subsize > 1, XBar&R, XTilde&R or XBar&S, for subsize = 1, RX, XMR(2) or XMR(3)) and the number of pieces of data anayzed. And of course which pieces of data. If you have subgroup size 1 in DataPage and are using the RX chart, and subgroup size 1 in DataView and are using the XMR(3) chart then you will get different values for the calculations. This is because Cp and Cpk use estimates of the standard deviation that are derived from the current chart selection and subrouping settings. The values Pp and Ppk should agree irregardless of the chart selected as it uses the population estimate of the standard deviation in the calculations. There is a discourse in the help file regarding the calculatoin of capability indices.


My understanding from the help file and OPTIONS - SPC CALCULATIONS - SUBGROUP PARAMETERS of Datapage that Pp and Ppk uses sample standard deviation and not estimate, therefore your values would be different from Cp and Cpk unless you change the setting for Cp and Cpk to use sample standard deviation from estimate standard deviation. Is this correct? Thank you.