I have a lengthy question, got a minute?

[Sl33stak]

OK, I got a letter from my boss, the letter explains how a customer of our is checking a part we make. Their method differs from ours, I'm going to give their method (verbatim), I was wondering if anyone can tell me if PC-Dmis is being used to check this part based on the description I give?

1.) Take 9 hits on Datum A (plane) est. best fit plane
2.) Measure shaft holes (B&C Datums) using 9 hits on each to establish 2 lines intersecting the A Datum. A PLANE is created from these two lines. This now creates 2 intersecting planes Plane A, and Plane BC. (Can this be done in PC-Dmis? Can we take two lines and construct a plane between them? (it looks like they are picking up B&C as cylinders right?)

Or does this look the be done in another software package?

The way I do it is. (this was written by the previous programmer, so I hope it's right!)

1.) 10 Hits on plane (Datum A) Level it.
2.) Pick up Shaft holes as circles (B&C Datums) using 4 hits.
3.) Create points that intersect from circles (B&C) to Plane (A)
4.) Contruct a line from B&C Points.
5.) That line is my rotation feature

Is this equivalent to their method? Input?

[Wes Cisco]

OK, I got a letter from my boss, the letter explains how a customer of our is checking a part we make. Their method differs from ours, I'm going to give their method (verbatim), I was wondering if anyone can tell me if PC-Dmis is being used to check this part based on the description I give?

1.) Take 9 hits on Datum A (plane) est. best fit plane
2.) Measure shaft holes (B&C Datums) using 9 hits on each to establish 2 lines intersecting the A Datum. A PLANE is created from these two lines. This now creates 2 intersecting planes Plane A, and Plane BC. (Can this be done in PC-Dmis? Can we take two lines and construct a plane between them? (it looks like they are picking up B&C as cylinders right?)

Or does this look the be done in another software package?

The way I do it is. (this was written by the previous programmer, so I hope it's right!)

1.) 10 Hits on plane (Datum A) Level it.
2.) Pick up Shaft holes as circles (B&C Datums) using 4 hits.
3.) Create points that intersect from circles (B&C) to Plane (A)
4.) Contruct a line from B&C Points.
5.) That line is my rotation feature

Is this equivalent to their method? Input?

Well I do not know that we can tell if it pc-dmis they are using or not with only the info you provided. But. . . ( there always is one, and sometimes it has more than one letter t)

1) You need to know if they are taking 9 hits to make a circle or a cylinder.
2) Are the shaft holes B & C perpendicular to Datum A?
3) If you measure each of the shaft holes as two circles each, you can then construct a plane from those circles. Pc-dmis wants at least 3 points to create a plane.
4) If the holes are perpendicular to Datum A, I see no real difference in using a plane or a line for rotation, but if the holes are not perpendicular then a plane could make some difference. (I wouild use a line like your program does.)
5) I must say I do not think using these holes to make a plane for rotation is good practice. You may well be dealing with a company that has a poorly trained CMM person. (Worse yet, they may be one of those people who just does't get it!)
6) Why was the letter sent? Do they want you to measure the parts the same way they do just on principle or are they expecting your results to match their results to a high degree? If the former then asking them some of the questions here should help you decide whether or not you should change. If the later, then you need to get into all the nitty gritty stuff, well beyond alignment, although alignment(s) are a biggie. Types of CMM/software, (even version level)/probe system, tip length, order of measuing features, methods of measuring features (number of hits, number of rows, etch), and it goes on and on. Even if you both have exactly the same equipement and even run the same program, you will not see complete exact agreement to the tenth between the two machines.

I think the motive thing is key. What are they looking for, (hopefully they know and are not just creating noise and confusion to hide their own incompetence!), then you decide how to go about giving it to them. HTH

P.S. Disclaimer, I am only half awake and need more coffee, so if I made typos or said something dumb, I will try to fix it soon.

[#2#]

Well, I don't know if it can be done in PC-DMIS. I read it as they are measuring B and C as cylinders - otherwise I can't see how they create a plane perp. to A.

My biggest question is how are they aligning? Lets say that A is up and B and C are not quite perpendicular. If they align by leveling A to z-plus and then by leveling their not quite perp. plane to x-plus or y-plus then they are not truly aligned to ABC. Leveling the not quite perp. plane will change the alignment so that A is no longer level to z-plus. They would be much better off leveling their created plane and then leveling A OR, level to A, create a line and rotate the line to the appropriate axis (as you do).

[Wes Cisco]

Well, I don't know if it can be done in PC-DMIS. I read it as they are measuring B and C as cylinders - otherwise I can't see how they create a plane perp. to A.

My biggest question is how are they aligning? Lets say that A is up and B and C are not quite perpendicular. If they align by leveling A to z-plus and then by leveling their not quite perp. plane to x-plus or y-plus then they are not truly aligned to ABC. Leveling the not quite perp. plane will change the alignment so that A is no longer level to z-plus. They would be much better off leveling their created plane and then leveling A OR, level to A, create a line and rotate the line to the appropriate axis (as you do).

You should only level once per alignment. Trying to level to two planes in one alignment will cause pc-dmis to escalate the hair pulling component by several orders of magnitude, not to mention it will likely give bad data. It is ok to level to plane A and then rotate to plane B, but it is not ok to level to both. HTH

[#2#]

You should only level once per alignment. Trying to level to two planes in one alignment will cause pc-dmis to escalate the hair pulling component by several orders of magnitude, not to mention it will likely give bad data. It is ok to level to plane A and then rotate to plane B, but it is not ok to level to both. HTH

Good to know! I've never actually tried it. I've always figured it would be trouble so have tried to avoid it.

[craiger_ny]

You can level as many times as you want in an alignment. The active level will be the last one called. The hair pulling will be the alignment commands between each level (the translate, rotate etc). There is nothing wrong with more than one level in an alignment (if there is a need) PCDMIS does not treat it badly and give you bad data because of it. You get bad data because you did not understand the ramifications, it does exactly what you tell it to do in the order you tell it to do it. In this case I have to stick up for PCDMIS, it does not give random ("bad") data, it does what you tell it to do, the bad data comes from not understanding what is going on when you do it.

Craig

[Sl33stak]

OK, some answers. (the Co. in Question is one of the "big three")

I am assuming they are measuring B&C as cylinders, these Cyls. ARE perpindicular to A. I guess I was thrown off by them using a plane where I am using a line for rotation. I was wondering if they were constructing a plane between the two cylinders, and if so...why? (i know you can do this in ProE)
but wasn't sure if Dmis or another CMM program could do it.

They just want to know how we are laying out the part, I guess to make sure WE are doing it right. But I feel bettr because since the first post I noticed they are dim. three features wrong....so they for sure aren't perfect! LOL.

[Wes Cisco]

You can level as many times as you want in an alignment. The active level will be the last one called. The hair pulling will be the alignment commands between each level (the translate, rotate etc). There is nothing wrong with more than one level in an alignment (if there is a need) PCDMIS does not treat it badly and give you bad data because of it. You get bad data because you did not understand the ramifications, it does exactly what you tell it to do in the order you tell it to do it. In this case I have to stick up for PCDMIS, it does not give random ("bad") data, it does what you tell it to do, the bad data comes from not understanding what is going on when you do it.

Craig

Craig, you may be right. I have only done this once, in one of my first programs and yes I certainly did not know what was going on then, but after I fixed the alignment it worked and I vowed to stick to 3-2-1. I have seen one to posts to this forum that were solved by eliminating a double level in the alignment as well. But in the example at hand, if I understand what you are saying,if they level to the Datum A plane, then when they level to the Datum B_C plane that replaces the previous level and they still need to rotate. Is that correct or would they retain the rotation from the Datum A level?

[#2#]

Craig, you may be right. I have only done this once, in one of my first programs and yes I certainly did not know what was going on then, but after I fixed the alignment it worked and I vowed to stick to 3-2-1. I have seen one to posts to this forum that were solved by eliminating a double level in the alignment as well. But in the example at hand, if I understand what you are saying,if they level to the Datum A plane, then when they level to the Datum B_C plane that replaces the previous level and they still need to rotate. Is that correct or would they retain the rotation from the Datum A level?

If I understand the way it would work... if A is leveled to z-plus then the B-C plane is leveled to x-plus and they are not perp. A will no longer be level in the x dirction. A will still be level in the y direction.

So, you would need to level B-C to x-plus (level in z and y directions) then level A to z-plus. Now A is level in x and y directions and B-C is level in z direction.

I think I would do a little extra work to be safe. I'd create a line where the two planes intersect. Then level A and rotate to the line.

[craiger_ny]

Craig, you may be right. I have only done this once, in one of my first programs and yes I certainly did not know what was going on then, but after I fixed the alignment it worked and I vowed to stick to 3-2-1. I have seen one to posts to this forum that were solved by eliminating a double level in the alignment as well. But in the example at hand, if I understand what you are saying,if they level to the Datum A plane, then when they level to the Datum B_C plane that replaces the previous level and they still need to rotate. Is that correct or would they retain the rotation from the Datum A level?

I am not speaking to this question I am speaking to the comment that you will get bad data. To be honest I do not understand Sl33stack's problem. I just want people to know it is perfectly valid to level more than once I do it frequently for odd ball datum scenarios, it works just as it should. The "bad data" in this case comes from the programmer. Say you level1 > rotate1 > translate1 > level2 > rotate2. All of your rotate translate operations after level1 and before level2 are done while in the first level scenario, anything after level2 is done under that leveling scenario and the last level called it the level that your program codes runs under until the next alignment call. That is all I'm saying. It has it's applications and works like it is supposed to. I do not know if Sl33stack's application requires it, I'm a little puzzled by his problem.

Craig

[Sl33stak]

I have no problem, I had questions. I wanted to know

A.) Is PC-dmis being used to check this part?

B.) If so, how(and why) do they go about using two cylinders tocreate a plane to rotate about?

C.) Is the method I am using equivalent to theirs (the customer)?

The customer just wanted to know how it was being laid out here, they gave no reason that I am aware of, no quality issues.

And the datum structure of this part is totally made for a SIMPLE 3-2-1 alignment. No need for dual leveling.

[#2#]

Going back and looking at your original post...

I believe your way is the way to do it. I would align that way. I would make one change though. I would take more than 4 hits on the two circles. Maybe it won't matter but I usually take 7.

[bw_bob]

This is not about the right way or wrong way to measure the part. We talk with the vendor or customer and come up with a inspection criteria that we both agree on. The CMM program is done the same way. If that does not happen you might find it hard for corralation and if there is a problem it just turns into a p!$$ing contest.My .02

[Sl33stak]

Thanks everyone, Goodluck...Do you determine number of hits based on Dia.? These are relatively small diameters, so I have been sticking with four hits due to that.

[#2#]

Thanks everyone, Goodluck...Do you determine number of hits based on Dia.? These are relatively small diameters, so I have been sticking with four hits due to that.

To some extent. I am in a position where speed and throughput are not critical. I believe the number of hits is personal preference and also determined by the machine and the operator's experience. In my experience with my machine I don't get good data with less than 7 hits. As I am not in a hurry I usually make it 11 for any hole under 1" dia. I will up it with the diameter over 1" to try to keep the hits a resonable distance apart.

With small holes it is very important that your hit vectors be very good. I would probably measure each of the holes manually and then go back and measure them DCC using the re-measure function. This would help to make sure your vectors are as perp. to the surface as possible.

Again, this is my personal experience and preference. Other's opinions and experience may dictate different things. Basically, it boils down to what you are comfortable with.