Paper ID #15834

Development and Evaluation of a Computer Program to Assess Student CAD

Models
Dr. Steven Joseph Kirstukas, Central Connecticut State University
Steve Kirstukas is an Associate Professor at CCSU, where he teaches courses in solid modeling, MATLAB
programming, and engineering mechanics. He is exploring the use of computer-aided assessment of CAD
files to give consistent, accurate, and quick feedback to students. He has degrees in civil and mechanical
engineering, with a Ph.D. from the University of Minnesota. Steve has worked in industry as a civil
engineer, software developer, biomechanics researcher, and mechanical design engineer.

American
c Society for Engineering Education, 2016
 Development and Evaluation of a Computer Program
to Assess Student CAD Models

Abstract

This paper presents a computer program that can automatically assess the geometry and
changeability of student solid models. The program is launched from within the CAD software
package and a graphical user interface is presented so that the user has the opportunity to set
values for twelve different deduction categories. The program compares one or more student
model files to a single “gold standard” model. The program allows for the assessment to be
thorough, repeatable, non-subjective, and very fast. It can process 20 or more models per minute
and generates output very similar to what would've been generated by hand. The program has
been used to re-grade exams from the Spring 2015 semester; when computer-graded scores were
compared to instructor-graded scores, a strong correlation was observed between the two scoring
systems.

Introduction

At Central Connecticut State University, all engineering and engineering technology students
learn 3D solid modeling in a single course that covers part-making, formation of parts into
assemblies, and creation of layout drawings. We approach part model creation as an opportunity
to participate in open-ended engineering design: it’s one of the few times in the engineering
curriculum in which traditional mathematical problem-solving plays a limited role in engineering
design. Various different strategies can create acceptable solid models. For example, something
as simple as a washer can be modeled by 1) extruding two concentric circles, or 2) revolving a
rectangle about an axis.

When the modeled part is complex and utilizes multiple constrained sketches, extrudes, revolves,
holes, mirrored features, patterned features, etc., many of which may depend on one another,
evaluating the quality of the model becomes a time-consuming task1 and one that can be affected
by fatigue and influenced by subjectivity. It is important that the model not just reproduce the
required geometrical shape and size but that it be changeable in the future in a manner that
honors the design intent2. Because the creation of changeable models requires some planning2,
modelers are faced with the seemingly opposing concepts of producing a given geometry quickly
and producing a high-quality changeable model3. High-quality models are easy to change and
require little time to revise4. The development of quality solid models should be a principal goal
from the start of instruction4. Kirstukas5 proposed that design intent can be inferred by careful
inspection of the provided part drawing. If the drawing contains the necessary and sufficient
information to produce a unique solid model, then the solid model should utilize all drawing
dimensions, leave out no dimensions, and repeat no dimensions. Thus, the set of dimensions in
the model can be compared to those in the original layout drawing to form part of a quantitative
assessment of model quality.

Computer grading of model files has been a goal for a long time, in order to reduce the total time
needed by the instructor to assess many student models, to reduce the time between model
submittal and feedback delivered to the student, and to quantitatively assess model quality.
Baxter and Guerci6 described the existence of a computer program written to automate the
grading of SolidWorks files. The program compared key data from the student file to that of the
correct file. However, details on the grading algorithm and results were not presented. Ault and
Fraser1 developed a program to automate the grading of Creo files. The program compared
information from the student file to that of the instructor file, such as model volume, the presence
or absence of critical dimensions, and the total number of extrude, revolve, hole, and pattern
features. The program did not attempt to evaluate model changeability.

Both of these efforts read solid model files and compare information in one model to that of
another model. Because file formats are usually proprietary and typically change after major
software releases, a vendor-supplied software library called an Application Programming
Interface (API) provides tools to query the solid model.

In this paper, a new computer-graded scheme is described. The previously described algorithms
are expanded to evaluate shape and size, orientation, modeling practices, and changeability. The
computer program is written specifically to compare Siemens NX solid model files, although
much of the algorithm is relevant to other CAD software packages that have well-developed
APIs.

The File Comparison Program

The computer program presented here is written in the VB.NET language (a form of Visual
Basic, an object-oriented computer language that is relatively easy to learn) and is compiled in
Visual Studio to produce an executable file. The program uses the Siemens-provided NX Open
API to extract information from NX model files. There are many sample NX Open based
programs included with the NX CAD package, and the internet has a wealth of resources for
learning more about the potential of user-built NX Open based programs. The websites
www.nxjournaling.com and www.eng-tips.com provide additional sample programs, example
code, and a forum from which clues, hints, or complete solutions to specific issues can be
obtained.

The computer program compares one NX solid model file to another. When one model is the
instructor’s “perfect” model and the other model is the student’s model, the output of the
comparison can be used to assess the student model.

The file comparison program is launched from within NX and a graphical user interface (GUI) is
presented. The graphical user interface (Figure 1) is divided into three main groups —
Deductions, File Selector, and Options — for which the user may alter default inputs. Below
these three groups, there is an area for which output text may be displayed. The text output box is
stretchable and can be enlarged to fill the screen. In addition, there are two buttons, one for
clearing the text output box and one for initiating the file comparison.
 Figure 1. The user interface of the computer program.

A Deductions group lists 12 categories of issues that are common in beginning student models.
For each category, point deductions may be individually set. These 12 deduction categories are:

Units: The first decision the modeler must make is often the unit convention. In our CAD
software, length units are typically in either millimeters or inches. The correct choice is
important, as it is not easy to convert a model from one system of units to the other.

Unconstrained Sketches: Every sketch is expected to be fully and properly constrained with no
auto dimensions.

Auto Dimensions: A non-fully constrained sketch will generally contain one or more auto
dimensions. Auto dimensions are not locked down and indicate that the sketch can be deformed
or broken by dragging a vertex. This deduction applies an additional penalty on non-constrained
sketches.

Banned Constraints: Students are encouraged to make generous use of the geometric constraints
of horizontal, vertical, coincident, concentric, equal length, equal radius, tangent, point-on-curve,
midpoint, and a few others. These all lead to changeable models that incorporate desired design
intent. However, some beginning modelers, in an effort to eliminate auto dimensions, set certain
points as “fixed”, or line segments at “constant length” or “constant angle”. These three
constraints make the model unchangeable and are not to be used. Any sketch containing one or
more banned constraints will additionally be classified as a non-constrained sketch.

Missing/Repeated/Unwanted Dimensions: The modeler’s task is to transfer the design intent as

expressed in the layout drawing into the model. If the layout drawing contains the minimum
information to fully understand the part, there will be a one-to-one correlation between
dimensions in the drawing and dimensions in the model. If the drawing contains extra
dimensional data, then the burden is on the modeler to decide which dimensions are to be left
behind, or set as reference dimensions.

Unused Sketches: Beginning modelers sometimes create extraneous sketches that are not used in
any other features. These may contain some model dimensions and may be partially or fully
constrained. If these sketches were never used in features such as extrudes, revolves, or holes,
they don’t contribute to the solid model and the modeler should have deleted these before
submitting the model.

Non-United Bodies: In our CAD software, if a part is composed of non-united bodies then
operations such as placing holes through adjacent bodies, or placing edge blends on an edge
shared by the bodies, do not go as expected. Adjacent bodies should be united into a single body.

Shape and Size: The student model in its submitted unaltered form (static configuration) should
have the shape and size specified on the layout drawings.

Orientation: Models must be constructed with the goal of being used to create a layout drawing.
The placement of a model upon the world coordinate system should result in the desired
isometric orientation.

Absent Solids: Outside of the first couple of weeks, students construct solid models. If a solid is
expected but not produced, then models would lose these points, as well as deductions based on
shape and size, and orientation.

Negative Dimensions: The typical extrusion is created by sweeping a selected region in a

direction perpendicular to the region. An equivalent solid could be produced by extruding
material in the opposite direction by a negative distance. This is confusing and can be avoided. In
our CAD software, there is a “reverse direction” button so that there is never a need to use
negative numbers.

Missing Drawing Dimensions: While the program is primarily developed for evaluating sketches
and solid models, it can assist with the tedious evaluation of rather complex layout drawings. It
can verify whether the critical dimensions of the original drawing have been incorporated into
the final layout drawing.

Deduction Percent: Early in the semester when models depend on a single sketch and contain
just a few numbers, the values associated with the deduction categories would typically be
higher. Later in the semester when models are more complex, the deductions may be scaled
down. To reduce or eliminate the need for changing 12 deduction values, a 13th item has been
introduced. When deduction percentage value is 100, then a 2-point deduction would decrease
the total score by two points. If the deduction percentage value is reduced to 50, then the same 2-
point deduction would decrease the total score by 0.50  ×  2 = 1, or just one point.

The remaining two input groups are the File Selector group and the Options group. The File
Selector radio button group allows for the comparison of a gold standard model to a single
student model, or to all student files in a directory (class assessment mode). Single file
comparison is used primarily for debugging and testing. It is envisioned that at some point,
students will have access to the program and can use the single model option to compare their
model to the gold standard model. Finally, there is an inactive (not yet implemented) option to
permit the comparison of any number of selected files to the gold standard model.

The Options group allows the instructor to customize operation of the program for the models
being compared. A text field input is available to accept a list of comma-separated numerical
values that are optional for both the model and the drawing. For instance, if a value was on the
original drawing merely to enforce symmetry, the student may have been encouraged to use a
midpoint constraint instead. So that a model with or without the centering value would be graded
the same, that value could be input as an excluded value. Another common excluded value is
360, commonly seen in revolved parts or in rotational patterns. In the example of the washer
stated previously, one modeler chose to revolve a rectangle about an axis through 360°. But 360
doesn't have to be in the model and it would not be in the model if a second modeler chose to
extrude two concentric circles. Similarly, if there was a 45° chamfer of size 2, another modeler
may select an equivalent symmetric 2-millimeter chamfer and the 45 would never appear in the
model. In these examples, 45 and 360 are angular dimension values that could be specified as
“excluded values”. There is no penalty assessed if the model or drawing contains an excluded
value. It exists to avoid deductions for otherwise optional values.

If the student model layout drawing is not present or not of interest, the instructor would turn off
Verify Drawing Dimensions, causing review of the student’s drawing to be skipped.
Additionally, it sets the value of the deduction for Missing Drawing Dimensions to zero. This has
no effect on the scoring but is done to clarify that the category is inactive. Turning this option on
causes the value of the deduction for Missing Drawing Dimensions to be reset to the default
value.

If the option Run In Fast Mode is turned off, the program will show each processed model from
an isometric perspective, display its name, and pause until the user wants to resume. This allows
the user to get a qualitative sense of the quality of models that are being analyzed.

The last option Display Verbose Output allows for displaying a lot of information during model
evaluation to the text output window. This is useful mainly during debugging.

Program Initiation and Execution

When the instructor presses the Compare Models button, the program obtains the name of the
gold standard file and either the name of the one student file or the location of many student
files, as set by the File Selector group.
The program analyzes the gold standard model, evaluating all sketches and features, and stores
the key numerical values. Geometrical information such as the bounding box height, width, and
length is obtained. Additionally, the program extracts surface area, volume, and moment of
inertia about each of three axes.

The program then does a similar analysis of one or more student models. Every sketch in the
model is examined, whether external or internal. If a sketch was created but never used by any
other feature, it is a sign of sloppy modeling. An unused sketch should have been deleted by the
modeler. In these cases, the program counts the number of unused sketches, but does not further
penalize the model if those sketches are not fully constrained, etc. Sketches that are used by other
features are checked that they are fully constrained with no auto dimensions. If a sketch has auto
dimensions the program maintains a count of the total number of auto dimensions.

There are several geometrical constraints that students are taught to avoid because those
constraints limit the easy changeability of the model. These “banned” constraints (fixed, constant
length, and constant angle) are never needed and should not be used. The number of banned
constraints is counted.

If the part file also contains a layout drawing, then the list of drawing dimensions is stored. The
orientation of each dimension is also examined, as students are asked to always present
dimensions with horizontal orientation.

The set of excluded values entered in the GUI’s input text box are subtracted from the gold
standard model dimensions to produce a critical list of model dimensions. The list of student
model dimensions is compared to the list of critical model dimensions. The student model
dimensions that are not in the critical model list are categorized as unwanted dimensions. Items
from the critical dimension list that are not in the student model are categorized as missing
dimensions. Finally, the program examines the list of dimensions in the student model and looks
for repetitions (repeated dimensions). Unless a particular number was on the original drawing
more than once, it should not appear in the model more than once. There should be a one-to-one
correspondence between numbers in the original drawing and the numbers in the model.

Surface area and volume values are compared to see if the student part is the same shape and size
as the gold standard part, regardless of orientation. If the student part has correct size and shape,
then its orientation is verified by comparing the moment of inertia components and the bounding
box dimensions. However, moment of inertia components are unit-dependent. If the student used
incorrect units (for example, inches instead of millimeters), then direct comparison of inertia
components is invalid. Bounding box dimensions are independent of units and can be used to
crudely verify orientation provided that the model has dimensions where bounding box height,
width, and length are significantly different from each other.

However, if the part has a significant missing or incorrect feature but has otherwise correct
orientation, use of the moment of inertia components or bounding box dimensions may lead us to
an incorrect conclusion regarding orientation.
Extraction of Model Dimensions

A significant portion of the score depends on the total number of missing dimensions, repeated
dimensions, and unwanted dimensions. The dimensions of the model must be extracted from the
total list of expressions in a model. An expression in its simplest form could be something like
“p1=5”. The left side of the expression is a parameter name, in this case “p1”. The right-hand
side of the expression is a text string that may appear to be simply real numbers or they can be
rather complex mathematical expressions that can reference other defined parameters, for
instance “p7 = 17 – (p2 + 23)”. In this case, the program needs to extract all of the unique
numerical values on the right hand side so the p2 is disregarded and the 17 and 23 are stored. So
it’s not a trivial task to extract numerical values from an expression. The program must be able to
decipher algebraic equations involving numbers, parameter names, parentheses, and
mathematical symbols.

In the right-hand side text string, all mathematical operators are replaced with commas. Each
individual text string in the comma-separated text string is analyzed. If an individual text string
starts with “p” the program assumes that it has identified a reference to another parameter name
and disregards it. If an individual text string is a “2” or “0.5” and there is more than one
individual string, then that numerical value is disregarded. An expression such as “p7=50/2” is
commonly seen in various situations. In symmetrical components, to achieve a final width of 50,
it can be useful to perform a symmetrical extrude of “50/2” (rather than 25) in each direction.
The component will have a final width of 50, and the 50 is incorporated in the model by allowing
the CAD program do the math. The program would process the right hand side of “50/2” by
retaining the value 50, and disregarding the 2. Another situation where a divisor of 2 is seen is in
a parametric equation for centering. While a midpoint geometric constraint would be preferred, a
parametric equation is considered acceptable provided that it does not introduce any extra
dimensions into the model. Finally, while it would always be preferable to see circle diameters
specified directly with diameter dimensions, it would be acceptable to specify a radius as a
diameter value divided by 2. Either way, the critical diameter dimension is in the model and can
be quickly located and easily modified.

All dimensions in sketches are stored. All non-zero dimensions in other features are stored.
Zeroes in sketches indicate the need for more geometrical constraints. The number of negative
values in the student model is counted.

Calculations of part volume, surface area, and moment of inertia pertain to a single body. If the
model is composed of various non-united bodies, these bodies are temporarily united for
purposes of evaluating shape, size, and orientation.

Comparisons of Computer Grading to Instructor Grading

In Spring 2015, a final exam based on a “back-of-envelope” sketch (Figure 2) was given to a
section of 19 students. The students were asked to construct a model and make a layout drawing.
 Figure 2. The part to be modeled during the Spring 2015 final exam.

This model requires three sketches and three extrudes to form the main geometry. It is not
immediately apparent which sketches and features must be created before others. It is up to the
student to study the drawing, understand dependencies, and formulate a modeling plan.
Additional sketches are required to place the holes and to create the pattern of recessed triangles.
Careful selection of the order of events can lead to a simpler model. For example, if the edge
blend operation along the curved edge is done before the two top holes, then it is a simple one-
click operation. But if done after the holes, then the curved edge has been broken into three
segments and creating the edge blend becomes a clumsier operation.

The final exam models were hand-graded by the instructor at the end of the Spring 2015
semester, before the first line of computer grading code had been written. The instructor
reviewed each model by opening it, turning off all features, then sequentially turning the features
on in chronological order and recording the number of occurrences of specific issues in a
spreadsheet-based scoring system (Table 1). The right-most column shows how a specific
student model was evaluated.
 deduction max%number number%of%observations
Exam%3%Deductions
value of%deductions in%a%student%model
Model&missing&dimensions&from&original&drawing 1
Model&uses&numbers&not&from&original&drawing 4
Model&has&redundant&dimensions 4
Total&number&of&dimensional&issues&in&the&model 2.5 12 9
Unused/unneeded&sketches&or&datum&planes 1 3 0
Issues&in&part&tree&(e.g.,&exclamation&points) 3 3 0
Model&is&unnecessarily&complex 3 1 1
Sketch&not&fully&constrained 6 3 0
Incorrect&feature&or&dimension 3 6 1
Other%comments: missing&dimensions:&76
repeated&dimensions:&55,&8,&46,&24
unwanted&dimensions:&30,&22,&260,&12
incorrect&radius&on&front&face
Exam%3%Model%Score 72
Table 1. The grading spreadsheet showing the evaluation of one student model.

For a given item, there was a limit on the number of deductions. Using this information, the
spreadsheet computed a final score:
𝑠𝑐𝑜𝑟𝑒   =  100  – 7689 𝑖𝑡𝑒𝑚𝐷𝑒𝑑𝑢𝑐𝑡𝑖𝑜𝑛6 Eq. 1
where,
𝑖𝑡𝑒𝑚𝐷𝑒𝑑𝑢𝑐𝑡𝑖𝑜𝑛6 =   𝑑𝑒𝑑𝑢𝑐𝑡𝑖𝑜𝑛𝑉𝑎𝑙𝑢𝑒6  ×  𝑚𝑖𝑛(𝑛𝑀𝑎𝑥6 , 𝑛𝑂𝑐𝑐𝑢𝑟𝑎𝑛𝑐𝑒𝑠6 )

For the student model shown in Table 1, the score is thus computed as,
𝑠𝑐𝑜𝑟𝑒   =  100 − 9(2.5) − 1(3) − 1(3)   =  71.5

Computer Assessment of Student Work

The same deduction values displayed in the user interface in Figure 1 were used, with two
changes. First, as we are only interested in examining the model and not the layout drawing, the
option Verify Drawing Dimensions was turned off and thus the value of the Missing Drawing
Dimensions deduction was set at zero. Because orientation of the part was not pre-specified and
because more than one view of this symmetric part makes an acceptable isometric orientation,
the deduction value for Incorrect Orientation was also set to zero. The same student model that
was evaluated manually in Table 1 was re-evaluated by the computer program (Table 2).

Your  model  has  the  correct  shape  and  size.  

Your  model  is  missing  2  dimensions  from  the  original  drawing:  40,  76.  
Your  model  has  10  repeated  dimensions:  8,  12,  20,  22,  22,  22,  24,  46,  55,  130.  
Your  model  has  2  dimensions  that  were  not  from  the  original  drawing:  30,  50.  
Use  fewer  numbers  and  more  geometric  constraints!!!  
Model  Score:  72  
Table 2. Output of the computer program after analyzing the student model.

The two schemes produced virtually the same score. This is partly because there is some
similarity in the algorithms that produce the scores. The computer program employs a score
calculation similar to Eq. 1, except that there are now 12 deduction categories and Deduction
Percent is incorporated. But this exact correlation is not expected. Note that in the manual
grading scheme, some dimension issues were missed due to sloppiness of the instructor. Others
were categorized differently (22 is an unwanted dimension in the manual evaluation and multiple
instances of 22 are categorized as repeated dimensions in the computer evaluation).
Miscategorization of dimensions does not affect scoring, as missing/repeated/unwanted
dimensions all carry the same deduction values.

A high-quality model will respond to changes in its dimensions in ways that produce expected
outcomes, without surprises. Changeability can be assessed by attempting to change a few pre-
determined dimensions in a model2. For many semesters, the instructor tested the changeability
of each student model by changing a few dimensions and observing what happened. It is still
done occasionally, but only for the benefit of the students to demonstrate that models that lack
the one-to-one correspondence of original drawing dimensions to model dimensions will have
issues when the model needs to change. In the Spring 2015 model, we could modify the height,
width, and depth of the part by changing three dimensions: 135 → 160, 100 → 120, and 24 →
30. To change these dimensions in the model, we just locate the first chronological instance of
the desired number and update it. Because the model was designed with a set of unique
dimensions, if we can locate the desired number, we know it is the number to change.

As noted in the computer output of Table 2, the student model has the exact static geometry as
desired. However, due to the lack of one-to-one correspondence of original drawing dimensions
to model dimensions, it does not handle change very well. In Figure 3, we see the instructor’s
model on the left (which was identical in appearance to the as-submitted student model). In the
center, we see the instructor’s model after the three dimensional changes. It honored the design
intent in the original drawing, and just got larger. On the right, we see that the student model
developed some issues during resizing. The edge blend was lost, the pair of top holes became un-
centered, the radius of the curved surface increased, the gap between the interior side walls
increased, the dovetail width increased, and some extra geometry appeared in the back of the
part.

Figure 3. The desired initial model, the instructor’s model after three dimensional changes, and
the student model after the same dimensional changes.
Computer Evaluation of an Entire Course Section

The computer program was used to evaluate the models of an entire section of 19 students from
the Spring 2015 final exam. Results were compared to the previous instructor evaluations of
these same models (Figure 4). Data points that lie on the diagonal line indicate perfect agreement
between instructor and computer. Data points significantly off the diagonal trendline could
indicate a need for tweaking the deduction values or otherwise improving the algorithm. If a data
point was significantly above the diagonal trendline, then it could indicate favoritism. If a data
point was significantly below the trendline, then it could indicate a negative attitude toward that
student.
Instructor  vs  Computer  Assessment
100

90

80
Instructor-­‐Graded  Score

70

60

50

40

30

20
20 30 40 50 60 70 80 90 100
Computer-­‐Graded  Score

Figure 4. The scoring systems compared to one another.

The coefficient of determination (R2) can provide some quantifiable information on the
correlation between computer-graded models and instructor-graded models. When all 19 student
models are evaluated together, R² = 0.65.

The method is best at differentiating fair and good models from excellent models. Student
models that have major issues can actually score lower than an empty model if they have enough
non-constrained sketches, missing/repeated/unwanted dimensions, etc. When an empty model
(nothing but a coordinate frame) is compared to the Spring 2015 gold standard, it scores 38. The
lowest scoring student in Figure 4 scored 28 in the computer-based assessment. Even though the
student demonstrated some capability in various modeling strategies and captured some aspects
of the correct geometry, the program has trouble determining this. If we leave out the one outlier
student score, R² = 0.86, demonstrating a strong correlation between the scoring systems.

Program Limitations and Future Work

Outside of volume and surface area comparisons, very dissimilar models can appear similar
when reduced to a list of dimensions. For instance, a cylinder with diameter 50 and length 200
will have the same critical dimensions as a square plate that is 200 on each side and 50 thick. The
deductions for shape and size as currently set are relatively small, so an inappropriate model with
the expected dimensions could score well.

The program does not currently monitor the number of instances of certain non-pattern based
features. For example, if an R4 edge blend was specified at three places, but it was used only
once (perhaps possibly even on the wrong edge), then the “4” would be extracted from the
student model and compared to the “4” in the gold standard model. The lack of the other two
edge blends would be ignored, as well as the possible incorrect edge election. Comparisons for
shape and size may detect the error, but the default deductions are relatively small.

The deduction values could be adjusted to increase the correlation between the quality of the
model and its score. No systematic attempt was made at finding the set of deduction values that
would minimize the difference between the scores of the human grader and the computer.

The current version of the program has processed many different models of various complexity,
with features including revolves, extrudes, edge blends, chamfers, various hole types, mirrored
and/or patterned features, and internal and external threading. Furthermore, versions of each of
these models have been produced by 40 or more different students. Although students sometimes
use non-taught and unexpected features of the CAD program, the program has been expanded to
handle these cases. The computer program now successfully handles all models that it has
encountered. Every large set of new models has the potential to reveal some weakness in the
program or room for improvement.

Conclusion

Hundreds of hours have been invested in the creation of the program in an effort to save
hundreds of hours of grading time in the future. We're already reaping benefits, as the computer
program allows for the grading to be thorough, accurate, repeatable, non-subjective, and very
fast. It can process models in seconds and generates output very similar to what would've been
generated by hand. We feel that prompt feedback is essential for students to improve their
models and make solid modeling more of an engineering design task and less of an art project.
We find that the computer, unlike the instructor, never suffers from fatigue, does not miss
anything that it was programmed to look for, and is not subjective in the assessment of students.
It treats all students identically whether they personally have a good relationship or a not-so-
good relationship with the instructor.
While the computer code that has been written and evaluated is specific to one particular CAD
software and API (Siemens NX), much of the algorithm can be transferred to other CAD
software packages that have well-developed APIs.

Bibliography
1.   Ault, H. K., & Fraser, A. (2013), A Comparison of Manual vs. Online Grading for Solid Models Paper
presented at 2013 ASEE Annual Conference, Atlanta, Georgia. https://peer.asee.org/19045

2.   Ault, H.K. & Giolas, D.T. (2005), An investigation of solid modeling practices in industry, Engineering Design
Graphics Journal, Vol 69, No.1, pp. 34-43.

3.   Ault, H. K., & Bu, L., & Liu, K. (2014), Solid Modeling Strategies – Analyzing Student Choices, Paper
presented at 2014 ASEE Annual Conference, Indianapolis, Indiana. https://peer.asee.org/23026

4.   Otey, J. M., & Company, P., & Contero, M., & Camba, J. D. (2014), A Review of the Design Intent Concept in
the Context of CAD Model Quality Metrics, Paper presented at 2014 ASEE Annual Conference, Indianapolis,
Indiana. https://peer.asee.org/19992

5.   Kirstukas, S. (2013). A Preliminary Scheme for Automated Grading and Instantaneous Feedback of 3D Solid
Models, Proceedings of the Midyear Conference of the Engineering Design Graphics Division of ASEE, pp. 53-
58.

6.   Baxter, D., & Guerci, M. (2003). Automating an Introductory Computer Aided Design Course to Improve
Student Evaluation, Paper presented at 2003 Annual Conference, Nashville, Tennessee.
https://peer.asee.org/11479