Absolute Accuracy Value for Large Assemblies

[Chris3]

For those of you who have large car type assemblies, what absolute accuracy value are you using in your start part?

[tbraxton]

I would not use a single value for absolute accuracy. I would suggest abs accuracy values based on the manufacturing process/tolerances of the parts. You will also need to manage the values in the context of data sharing features as the values should match between dependent models.

If you use a single value, it will be tighter than needed for many of the parts. The downside to this is a potential performance hit during regeneration. If you are not worried about the HW loading, then using a single value will work. I have no idea how much of a burden it is when tightening the abs accuracy by say 2x.

If you use a single value, you will need to use the tightest value of all of the components. An injection molded lens for the lights will need a much tighter value than the stamped body panels as an example.

[rreifsnyder]

The model accuracy does not have anything to do with manufacturing the part, It has to do with the number of decimal places Creo uses in computing the geometry. of the part. You're correct that setting everything to the tightest setting would increase the regeneration times for the models but with today's computers I don't think it would be a major issue. Also, in the past at least, things like interference checks could be unreliable if the accuracies of the parts were different. 

In your example you compare a molded lens to a body panel implying that since the body panel is stamped that it could use a looser accuracy value but in fact it could be the opposite. The complex geometry of a body panel often means that concave surfaces need to flow into convex surfaces with bends between some times. Basically the math to make sure the surface edges come together correctly means that it likely needs tighter accuracy.

[tbraxton]

I am offering observations of how absolute accuracy affects the design and manufacture of components. I am offering this info for consideration for how one may choose to implement absolute accuracy. IME the same model accuracy is not used across all part designs in a car, airplane, etc.

I work across many industries including transportation and aerospace in mixed CAD design environments. The model accuracy value is absolutely without question directly related to successful manufacturing of parts when the CAD model is used to program tool paths etc. I have had to fix accuracy problems that are preventing the creation of parts.

I also have experience with the design of Class A surfacing of exterior panels used on ground vehicles and aircraft using Creo and ICEM surf.

The assignment of the value used for absolute accuracy should in fact be related to the manufacturing process and the permissible tolerances on features. As an extreme example I have start parts that are used in sensor design for MEMs systems and the model units are angstroms. These sensors are integrated into much larger products. The MEMS start parts do not use the same abs accuracy as the start part for a sand cast engine block.

[TomU]

Also keep in mind that it is very important to match accuracy between models and their assemblies if using any of the data sharing features (copy geom, etc.).  This is where having a single value everywhere is helpful.  We don't design cars, but we do have all of our start parts (templates) set to the same absolute value.

[tbraxton]

@TomU 

I agree with you completely on matching abs accuracy in this context. If you do not do this, it will be a painful path to work on the designs.

I did note in my first response to the OP:

"You will also need to manage the values in the context of data sharing features as the values should match between dependent models."

[Chris3]

To restate I am curious about your start parts. I understand that different parts may have different accuracies but everyone has start parts.

[Chris3]

PTC wants to close out this thread but no one actually offered to state what they are using for start part accuracies. Any last minute takers before we close as unresolved?

[StephenW]

Our assembly start part absolute accuracy is 0.008 (inch units). It's our basic assembly start part, no difference for large/small.  I don't know why .008 was chosen unfortunately.

[Chris3]

PTC is again pushing me to accept a solution so I will mark @StephenW response as the solution because he provided one data point although I was hoping to get others.

[StephenW]

By no means do you have accept any answers as a solution. If you keep the post active, you will likely get a few more answers. To be honest, your question doesn't have a solution. You are just gathering data. 

[Chris3 · Accepted solution]

Some other data points I have for others in the future:

PTC's absolute in start part uses: 0.00039 (converted from .01 mm)

In this thread, FPD Company says theirs is .0001 (smaller than car sized assemblies)

In this thread, @kdirth  and @tbraxton say they are suing .001 mm (no idea the size of their products)

I spoke with a large equipment MFG at the last PTC user meeting and they said they were using .0004 (in)

Our largest products are car to truck sized and I am leaving towards .0004 (in)

[StephenW]

You should mark your post as the answer or leave it as open. You post has much more data than my one random answer that likely doesn't correspond well with what you are asking for.

I have a hard time comparing my assemblies to a car/truck. My part of the job is mostly the big dumb steel that interfaces with the outside world. But within my assembly are other groups stuff that is much "finer". I pay little to no attention to what they do internally and I try my best to strip that out of my model, aka simplified reps or envelopes.  My assemblies are physically large, relative to a car...my largest project is just over a million pounds, ~700 inches tall (~58ft or ~18 meters)

[Chris3]

Our products have lots of detail in the part levels and when you combine those to a vehicle sized assembly large assembly management becomes very important. Up until this point we have stayed with relative accuracies because I don't have the internal funding to do a study on converting one of our large assemblies to absolute and then running relative performance testing. After talking to some folks at the last PTC User meeting though we are going to switch to absolute.

[StephenW]

I'll be completely honest, we "switched" from relative to absolute accuracy some time back. I don't even know if the user base was told. The start part was just different. I didn't notice any issues that I can connect to that transition.