Discussion of "Are We Training Our Novices Towards Quality 2D Profiles for 3D Models"

[Mark Szlazak]

Edit by Moderator: This topic was split from https://forum.freecadweb.org/viewtopic.php?f=36&t=36656

On the Role of Geometric Constraints to Support Design Intent Communication and Model Reusability

The assurance of model quality in parametric CAD implies that designers must build models that facilitate reuse while retaining their design intent when modified. In this context, a suitable selection of geometric constraints that operate within and between features is key to produce robust-while-flexible models which are a prerequisite for reusability. This paper introduces a new classification of 3D CAD model constraints that builds on the idea that making the meanings, similarities and differences explicit can result in better usage, making models robust to prevent catastrophic failures when edited as well as flexible enough to enable easy editing. Results from a preliminary user study designed to validate the significance of the new classification are discussed. Simple exposure to the proposed classification appears insufficient to validate effectiveness. Future work on educational approaches that focus on the challenges of selecting an effective set of constraints for particular reuse scenarios is suggested.

On the Role of Geometric Constraints to Support Design Intent Communication and Model Reusability .pdf

(648.84 KiB) Downloaded 121 times

Are we training our novices towards quality 2D profiles for 3D models?

In the history-based, feature-based, parametric CAD approach, 2D profile sketches are the basis for 3D models. Fully-constraining profiles is mandatory to create robust profiles. At present, neither CAD applications nor Model Quality Testing Tools usually check whether 2D profiles contain redundant constraints. Besides, our experience shows that novices tend to introduce redundant constraints. We hypothesize that 2D profiles over-constrained with redundant relations are more difficult to edit than those that avoid redundancies. In the present work―and as a first step to demonstrate this hypothesis―an experiment was conducted. Students of the subject “Graphics engineering” were taught on the creation of constrained 2D profiles. Then, they were asked two questions. On the one hand, novices had to identify and reason whether a simple given profile was fully-constrained, over-constrained or under-constrained. On the other hand, they had to identify and point out the types of the constraints. The results showed that in spite that novices received a specific training, roughly half of them failed to say if the 2D profile sketch was fully-constrained and which type of constraints it contained. Furthermore, the results of the second question revealed that more than the half of students did not recognize perpendicularity as a geometric constraint. As future work, we will try to demonstrate whether a reinforced training through simple exercises and a quick and effective feedback, will allow novices to improve the identification and removal of redundant 2D constraints when drawing 2D profile sketches (thus helping to produce robust profiles).

Are We Training Our Novices Towards Quality 2D Profiles for 3D Models.pdf

(246.5 KiB) Downloaded 118 times

[bejant]

At present, neither CAD applications nor Model Quality Testing Tools usually check whether 2D profiles contain redundant constraints. Besides, our experience shows that novices tend to introduce redundant constraints. We hypothesize that 2D profiles over-constrained with redundant relations are more difficult to edit than those that avoid redundancies. In the present work―and as a first step to demonstrate this hypothesis―an experiment was conducted. Students of the subject “Graphics engineering” were taught on the creation of constrained 2D profiles. Then, they were asked two questions. On the one hand, novices had to identify and reason whether a simple given profile was fully-constrained, over-constrained or under-constrained. On the other hand, they had to identify and point out the types of the constraints. The results showed that in spite that novices received a specific training, roughly half of them failed to say if the 2D profile sketch was fully-constrained and which type of constraints it contained. Furthermore, the results of the second question revealed that more than the half of students did not recognize perpendicularity as a geometric constraint. As future work, we will try to demonstrate whether a reinforced training through simple exercises and a quick and effective feedback, will allow novices to improve the identification and removal of redundant 2D constraints when drawing 2D profile sketches (thus helping to produce robust profiles).

Hmmm... looks like nobody used FreeCAD here (they should have)...

[bejant]

Section 2 of the "Are We Training Our Novices Towards Quality 2D Profiles for 3D Models.pdf" document states that university students are given a midterm test:

2 Experiment

One of the learning outcomes of the subject “Graphics engineering” (third course of “Bachelor’s Degree in Mechanical Engineering” at Jaume I University) is that students have to apply computer-aided three-dimensional modeling (3D CAD) for the resolution of graphic engineering problems. Throughout the course, students attend both theory and practical classes.

In the beginning of the course, students are introduced to the creation of 2D profiles following the first chapter of [20]. On the one hand, during theory classes, students receive knowledge about constraints. In addition, they learn to distinguish the differences between over-constrained and under-constrained profiles. On the other hand, during lab hours, students are introduced to the use of a 3D CAD application (SolidWorks®) and specific rubrics to convey quality criteria in Mechanical CAD Systems (assertions maps can be consulted in [11]).
Students are trained to create robust and flexible profiles through easy and simple examples

...snipped out of the quoted text by bejant...

In the first midterm exam of the 2017/2018 course, students who coursed this subject had to solve two questions related with constraints. In the first question, students were asked if the profile showed in Figure 2 was fully-constrained, over-constrained or under-constrained. The answer had to be reasoned.

20190708a.png (44.39 KiB) Viewed 3031 times

Given only the information in that image and without recreating a Sketch in FreeCAD:

Is it fully constrained or not, and how did you arrive at your answer?

I think the people grading the test (and also authors of the .pdf) arrived at the wrong conclusion.

[kisolre]

Assuming 5 coincidences (can't be visually checked from the picture). 3 perpendicularities mean fourth is also 90deg - we have a rectangle. Two equal sides - square. Size of square side fully constraints square dimensions. Coincidence with origin of coordinate system constraints the position and angle to coordinate system (if directional) fully constraints all.
And trying in FreeCAD checks out.

[chrisb]

If assuming that "coincident" means "all points coincident" and not only "one coincident constraint" then it boils down to the question what kind of perpendicular constraints we have here. If they are point-to-point then the calculation is as follows:

4 lines create 16 degrees of freedom (DOF)

They are consumed by
4 DOF by a double coincidence in the center
1 DOF equality
1 DOF dimension
1 DOF angle
9 = 3*3 DOF perpendicular

This sums up to 16 DOF which means it is fully constrained. The way of reasoning is rather efficient, because no further geometric understanding is necessary.

If the perpendiculars or the double coincidence mean something different then it will no longer be fully constrained.

[bejant]

The text I added to the upper right of the image reads: Don't these other 3 corners still need Coincident Constraints?
The text I added to the lower left of the image reads: Are 2 Endpoints Coincident to the origin, and also to one another?

The way the image in the test is drawn, I think there should be a Coincident Constraint applied at each of the other 3 corners. To me it also seems unclear if the Coincident Constraint at the origin is applied to 2 or to 3 Points.

20190709a.png (66.32 KiB) Viewed 3040 times

Based on the Coincident Constraint symbol at the Origin, in SolidWorks it looks like a Coincident Constraint symbol is shown where a Coincident Constraint is applied at a Point. Even so, their midterm test image wouldn't work for a FreeCAD test because Coincident Constraint symbols aren't shown in the Sketch so I think more information would have to be given to the people being tested.

[bejant]

Assuming 5 coincidences (cant be visually checket from the picture).

I agree that the image is unclear; to me it seems that the person taking the test can't tell exactly how many Coincident Constraints are applied.

[chrisb]

The way the image in the test is drawn, I think there should be a Coincident Constraint applied at each of the other 3 corners. To me it also seems unclear if the Coincident Constraint at the origin is applied to 2 or to 3 Points.

These are exactly the flaws I had seen. Especially the coincidence in the center is unclear. I'm not sure about the perpendicular constraints in other systems, and of course we don't know what was taught before to the students: In FreeCAD we have three variants of perpendicularity, namely point-point, line-point, line-line. Is this in other CAD systems the same?

[TheMarkster]

Should FreeCAD have some kind of coincidence constraint icon like this? I think it probably doesn't simply because of the clutter it would cause and the issue of overlapping icons is avoided. How about having the missing coincidence constraint highlighting found in the sketch validation tool automatically applied to sketches in sketch edit mode? But this might incur an expensive performance penalty, that I don't know. If so, then a toolbar action to highlight them on demand could be warranted.

[openBrain]

Should FreeCAD have some kind of coincidence constraint icon like this?

I'd really wish it as it. I'd like something very simple, like point changing its color or having an extra central black bullet, when a coincidence is set.