Assembly: Specifying type of fit and tolerance in assembly/joints

[kcleung]

I have heard that assembly and joint are undergoing a major overhaul.

Are there any plans to allow specification of type of fit (e.g. clearance, interference, transition etc) and tolerance for each joint? This would be useful for automated verification during the digital mockup so that things can be checked out before being made. I think this is done in other CAD/CAM software such as Solidworks.

If this is not yet planned, should I add a non-gating feature request in the issue so that this proposed feature is documented for future developers to pick up?

[gumyr]

The current focus of the Assembly system so far is to build the basics of the machine organization mechanisms.

If I understand your request correctly you'd like to be able to add a tolerance to the diameter of a shaft and hole and have the assembly system check that the fit matches the specified fit - i.e. for Fit.CLEARANCE the maximum sized shaft and minimum sized hole must never interfere.

This seems like a good direction to go but would add quite a bit of complexity to the system. There is a distance function that could be used to determine the minimum distance between two shapes. However, one would have to specify which sections of the part the fit applies to - with a bolt and nut the clearance only applies to the threaded part and not the head which could be tight against the nut. Any suggestion for how this might be done?

[kcleung]

The current focus of the Assembly system so far is to build the basics of the machine organization mechanisms.

If I understand your request correctly you'd like to be able to add a tolerance to the diameter of a shaft and hole and have the assembly system check that the fit matches the specified fit - i.e. for Fit.CLEARANCE the maximum sized shaft and minimum sized hole must never interfere.

That is what I meant @gumyr

This seems like a good direction to go but would add quite a bit of complexity to the system. There is a distance function that could be used to determine the minimum distance between two shapes. However, one would have to specify which sections of the part the fit applies to - with a bolt and nut the clearance only applies to the threaded part and not the head which could be tight against the nut. Any suggestion for how this might be done?

In the case of bolt and nut, it is my opinion that the fit between the two threads are not really clearance fit (c.f worm gear drive) as there meant to be friction between the two threads, otherwise the bolt can come loose as long as you take the wrench off!

Therefore, I think the fit between the two threads would be Fit.TRANSITION instead, which is equivalent to a push fit, as opposed to the press fit in Fit.INTERFERENCE

So in this situation, you can either put all contact surfaces between the bolt and the nut as a single joint with Fit.TRANSITION

Or alternatively, you can specify the thread surface and flat surface as separate joints, with the thread joint having Fit.TRANSITION and surface joint as Fit.INTERFERENCE

Though if I were to implement the assembly, I am more inclined to the former.

What do you think?

[gumyr]

Maybe the nut/bolt example wasn't a very good one. In general I'm concerned that it might be difficult for the user to specify a fit for difference sections of a part; however, this doesn't take away from the value of having a fit checker. The closest that currently exists is an intersection test on an assembly but that doesn't account for tolerances. I'll add this to the Assembly Roadmap - I'm okay will listing things that we don't have a clear idea of how they could be implemented.

Just brainstorming here - could there be a formal way to build tolerances into build123d models? Say the radius of a shaft being specified as 10mm +/- 0.02mm - interference checks could be done on different instances of the model with different tolerance values being applied.

[kcleung]

Maybe the nut/bolt example wasn't a very good one. In general I'm concerned that it might be difficult for the user to specify a fit for difference sections of a part; however, this doesn't take away from the value of having a fit checker. The closest that currently exists is an intersection test on an assembly but that doesn't account for tolerances. I'll add this to the Assembly Roadmap - I'm okay will listing things that we don't have a clear idea of how they could be implemented.

Just brainstorming here - could there be a formal way to build tolerances into build123d models? Say the radius of a shaft being specified as 10mm +/- 0.02mm - interference checks could be done on different instances of the model with different tolerance values being applied.

For formal ways to build tolerance, would it be a good idea to change a scalar specification (e.g. diameter) into a tuple {value, tolerance} with tolerance defaulting to zero if not specified?

This will be a good practice anyway if the user needs to turn the object into technical drawing.

So before the API are finalized in 1.0, would it be a good idea to transfer any scalar measurements from:

double value;

to the tuple / struct of:

struct value {
double value,
double tolerance
}

BTW anyone who is also a member of the partcad project: can you please comment the approach partcad wanna use?

[kcleung]

One more thing: Once we have enforced the { value, tolerance } tuple, the digital mockup simulator can then also take into account the data of the material of the two parts to determine thresholds of the transition and interference fits. We all know that the interference require for fits would largely depend on the hardness of the materials etc......

[jwagenet]

I don't know if Roger is changing the nature of joints with the Assembly overhaul, but as it stands none of the joint types contain any geometry information to assist with tolerance evaluation. In the same vein, with the exception of the *Hole objects, no other objects really have enough information in isolation to specify both the measurement and tolerance and have it certainly be relevant to the final part.

IMO, tolerance is not a property of either the geometry or the joint, but of the part itself, each consisting of at minimum list[Face, Edge, Vertex, etc] and (nominal_dim, lower_limit, upper_limit)

[kcleung]

I don't know if Roger is changing the nature of joints with the Assembly overhaul, but as it stands none of the joint types contain any geometry information to assist with tolerance evaluation. In the same vein, with the exception of the *Hole objects, no other objects really have enough information in isolation to specify both the measurement and tolerance and have it certainly be relevant to the final part.

IMO, tolerance is not a property of either the geometry or the joint, but of the part itself, each consisting of at minimum list[Face, Edge, Vertex, etc] and (nominal_dim, lower_limit, upper_limit)

But apart from holes. there are also other circumstances where tolerance needs to be tested for proper fit. A notable example is fixed joint with transition / interference fit is wooden interlock joints.

Therefore, it is in my opinion that all joints should have a FIT attribute and that all parts that are in contact within an assembly must be associated with a joint in order for the digital mockup function to properly do an assembly validity test prior to the physical manufacturing of the prototype.

I agree that the tolerance would be for the part itself. However, the FIT attribute will be in the joint. The assembly validity test would take into the FIT attribute of the joint, the tolerance of the parts if any, and the material of both parts in the joint, in order to run the validity test, right?

Furthermore, should we also be able to specify range of movements for a given joint? - say some hinge joints are 90 degree, some 180 degree etc?

[jwagenet]

I'm aware fits are not limited to holes; Tolerance Analysis can be one of the most important aspects of mechanical engineering. My point is Joints, as they exist, are effectively point objects with some special range of motion attributes (or none) and lack a lot of information required besides "fit".

For your wooden interlock example you would likely use a RigidJoint. How does the system know which pairs of faces you need to check the "fit" between? What if the tongue was constructed in such a way the width is indirectly specified and no tolerance is set for that feature? Fits which are affected by more than two members connected by more than one Joint? A dovetail joint with 3 contact faces instead of two?

I'm not saying Tolerance Analysis isn't useful, but I don't think "fit" checking is as simple as bolting on an attribute to a Joint.

Furthermore, should we also be able to specify range of movements for a given joint?

AFAIK CylindricalJoint already supports this

[gumyr]

I agree with @jwagenet here, a generalized "fit" system will be very difficult to develop as build123d (as with all CAD systems) doesn't put any limits on what is being designed. I agree that it would be a very powerful feature and one that should be in the Roadmap but not for the near term.

As for Joint movement ranges, they are already supported for both angular and linear ranges.