Point Clouds to Deliverable in AutoCAD Plant 3D

Scott Diaz kubit USA

J ohn Bunn -- kubit USA
Roland Legnon -- Autodesk, Inc.

PD2609 Point clouds continue to become a larger part of the plant engineer's and piping designer's workflow. This
class will take you through the entire process from a trip to the field for scanning, through the scan registration
process, and importing the scan data into AutoCAD Plant 3D software for design of needed deliverables. How do
real world and design world, and brownfield and greenfield meet in the Autodesk plant space? What are some of the
latest methods for going from cloud to intelligent plant object, detecting tie-in points, locating clashes, and creating
needed deliverables all within AutoCAD Plant 3D? Come to this class and find out!
Learning Objectives
At the end of this class, you will be able to:
Understand the entire journey from field scanning through design
Recognize the advantage of Autodesk point cloud capabilities within plant design
Learn Autodesk and Autodesk partner focused workflows for reaching needed deliverables in
AutoCAD Plant 3D
Use the tools needed to get from field to finish most efficiently
About the Speakers
Scott Diaz: Scott is managing director for kubit USA of Houston, Texas. Over the past four years he has
managed the US headquarters office in Houston, Texas representing kubit software directly for the
Americas while establishing a kubit reseller network throughout the US, Canada and Latin America.
Scott has a background in business management and marketing with an MBA from Lamar University.
Email: [email protected]
John Bunn: John is technical sales and support manager for kubit USA. He has represented kubit for
more than three years and is the top technical contact for the software in the Americas. john has more
than 14 years of AutoCAD experience as a user and professional programmer with specialty knowledge
in the industrial facility/piping design space.
Roland Legnon: Roland "Buddy" Legnon works on the Autodesk Plant Solutions Team. Buddy is an
AutoCAD Plant 3D technical specialist and has more than 20 years of user experience in AutoCAD. His
previous experience also includes time as a PDS and Autoplant administrator.





Point Clouds to Deliverable in AutoCAD Plant 3D
2


The Laser Scanning Process
Many people attending this class already have some exposure to the laser scanning process in
general. While most have seen a point cloud before, not everyone has had the opportunity to
understand the process of field data collection. This portion of the class will allow
readers/attendees the opportunity to understand the full surveying/scanning process in an
industrial setting.
For our trip to the field, our team is equipped with the Faro Focus 3D laser scanner, scanning
targets (spheres) and a supporting tripod. The device uses phase-based technology which
yields a high density result at a lower range distance. Density varies based on the user settings
but can range from about 15 million to more than 100 million points per scan position. The
scanner range varies based on environmental conditions but maximizes at 120 meters.
Typically we use high quality scan data within 30-40 meters of the scanner position. The time
per position varies based on the density settings and whether or not photographs are also
collected in addition to scan data.
Scanner Setup
Prior to setting up the scanner in the first position, it is recommended that the surveyor perform
a walkthrough of the area intended for data collection. The surveyor should assess the best
possible vantage points for reducing the number of overall scan positions, while still collecting
the needed data properly. The walkthrough will also help the surveyor visualize the area to be
collected from each intended position. This is important when assessing scanner target
positioning for quality registration results.

Setting Targets
Targets, as in traditional surveying, allow the surveyor to accurately tie each position to the next.
At least three targets should be visible to the scanner, per scan position. The user should also
Point Clouds to Deliverable in AutoCAD Plant 3D
3

assess whether these targets are visible from the next intended scan position. Setting out scan
targets is a crucial part of the initial scanning process for ensuring accurate data collection.
Our team used spherical targets provided by Faro Technologies. Alternate targets may include
checkerboard targets or known surveying coordinates (tied to the Plant or global coordinate
system). The Faro registration software (Scene) automatically recognizes these spherical
targets and uses them to triangulate and connect each position to the next.

Spherical targets should be distributed at various elevations and distances from the scanner
position. The surveyor should avoid setting targets in a linear fashion as this can impair the
ability to calculate quality triangulations.

Scanner Settings
Our surveyor set the scanner at 1/5 resolution which yielded 28 million points per scan position.
The scanner also has the added option of collecting photographs during the scanning process.
Each of the two scans would take approximately 3 minutes to complete.
Point Clouds to Deliverable in AutoCAD Plant 3D
4

Moving Scan Positions
When moving positions, it is important that the scan targets (spheres) which were visible in the
first position are visible in the second position. The scanner should offer a self leveling function
as well as keep the settings from the previous scan. Once the surveyor is comfortable with the
visibility and positioning of each position, the scanner is ready to collect data again. These steps
are repeated to each position and new targets can be added as well.
Scan Data Registration and Preprocessing
Each scan position was recorded to the internal SD card on-board the laser scanner. Once all
positions are complete, the surveyor is ready to take the collected raw data back to the office
and import it into the scanner registration software.
Data Import
For our Faro scanner, each raw scan position generated an FLS file. We will use Faro's Scene
software to walk through the scan registration process.
The team simply opens a new project inside of Scene and imports the collected FLS files. Upon
first glance, it is easy to notice that all scans are stacking on top of each other. Because the
positions are not yet registered, all scans are sitting at a 0,0,0 coordinate system and overlap
with many similar points.
Preprocessing
Our next operation is to choose the "pre-process" option inside of Scene. Here the team is able
to choose from a variety of scan registration methods. Since spheres were used in the field, this
option is selected. The software knows the exact parameters of these targets and will locate
them in each scan position. Based on the position of the spheres, the scan positions can be
correctly oriented to one another.

Detect Spheres
Point Clouds to Deliverable in AutoCAD Plant 3D
5

The registration process will create our registered and correctly oriented point cloud. The
accuracy report showed an overall standard deviation of 4 millimeters. When saving the project
in Scene, we find a set of registered FLS scan files (now in the revisions folder) along with a
project (FWS) file, which references each of the scan positions.


Applying Color
Once the registration process is complete, we can use the photographs collected to add color
values. This process took about 1 minute per scan. An RGB value is assigned to each point on
the cloud data.

Bringing the Data to Autodesk Software
Creating Autodesk Point Clouds
AutoCAD Plant 3D is used to import the native Faro scan data into the native Autodesk file
format. Since 2011, Autodesk has provided the PCG file format for managing large point cloud
data sets.
Create Point Cloud
To import the scan data into AutoCAD, we use the "Create Point Cloud" command found in the
AutoCAD ribbon under the INSERT tab. Users may also enter POINTCLOUDINDEX into the
Point Clouds to Deliverable in AutoCAD Plant 3D
6

command prompt of AutoCAD. This allows the user to choose from a variety of scanner
manufacturer file formats for import to AutoCAD. We find the Faro FLS and FWS formats for our
purposes. After selecting each file, the user can choose to save the file to the desired location
and the "indexing" process is started. This will convert the FLS files to the native PCG file format
for AutoCAD. About 2 minutes passes before the scans are fully converted into the PCG format
and ready to be "ATTACHED" into the AutoCAD drawing.
AutoCAD Settings
The POINTCLOUDATTACH command is used to insert the newly created clouds into AutoCAD.
Some settings to be aware prior to import are the following:
Many scanners collect data natively in meters. If working in imperial units, it is important
to first scale your cloud within the POINTCLOUDATTACH interface
Scan data does not function well, nor does it represent true intensity or RGB values in
2D Wireframe. A user should switch from this Shademode to any other available in
AutoCAD
Hardware acceleration should be turned on in AutoCAD. If a user is unable to turn on
this mode, the machine may be lacking in video card quality and/or be in need of video
card driver updates.

Now that the cloud data is successfully imported to AutoCAD, we can begin the process of
extracting information as shown above. How is this accomplished? Please look to the next
section "Shape and Data Extraction from the Point Cloud."
Shape and Data Extraction from the Point Cloud
This section explains how shape extraction algorithms, in combination with the proper catalog
can help a user model and extract piping and structural data from the point cloud.
Point Clouds to Deliverable in AutoCAD Plant 3D
7

Piping the Point Cloud
Extracting piping from a point cloud calls for needed pattern recognition algorithms, should be
catalog driven (based on standard piping design standards) and should create results which
follow the guidelines of the final design software (Plant 3D) while staying as close to the as-built
data as possible.

Shape Extraction Tools
Automatic shape extraction involves pattern recognition for standard
objects. The most common shapes extracted are primitive lines, polylines, boxes,
cylinders and cones. More specific extraction routines and algorithms may allow for
specific parts (fittings, beams, etc.) to be detected within the cloud.
The demonstration uses shape extraction algorithms from kubit to help close the gap between
points and intelligent model.
Defining Patterns for Extraction (Catalog Driven)
Shape extraction is most efficient when searching for the best-fit of a pre-defined 3D object.
What objects should we be looking for in the cloud? A piping designer should be looking for the
same patterns that he/she uses for new design. These come from standard piping catalogs and
specs already being used in Plant 3D.

In our demonstration, the kubit team shows how kubit PointSense Plant software will
automatically import all of the user's current catalogs/specs from AutoCAD Plant 3D into the
pattern recognition library. As a result, the software will now search for the same patterns that
the user has familiarity to and would use for new design.
Locate Patterns on Cloud Data
Using kubit's "Walk the Run" routine found in PointSense Plant software, the user is able to step
through a pipe run of a point cloud. Each component is detected along the line and the user can
confirm or override the pattern recognition. This process applies standard components derived
from the Plant 3D catalog to the point cloud data.
This process is meant to quickly and "locally" apply each pipe, elbow or inline fitting along a pipe
run. Local pattern recognition is the first step towards a fully modeled pipe run.
Point Clouds to Deliverable in AutoCAD Plant 3D
8


Apply Design Constraints
After local pattern recognition with the "Walk the Run" routine, we notice small gaps between
components and a lack of co-linearity between items. How can this be mended and why was
this process necessary?
The aim of pattern recognition software is to help close the gap between point clouds and
parametric objects. Design Software is made for Greenfield situations, not Brownfield. Such
programs typically
expect:
Connected objects,
Coaxial cylinder axis,
Coplanar cylinder axis,
90 angles and
Standard objects.

The following screenshots illustrate the need of coaxial and/or (at elbows and branches)
coplanar axes:

After local pattern recognition of each object along a pipe run (lower left), we have to resort to or
Point Clouds to Deliverable in AutoCAD Plant 3D
9

apply the constraints of our target software (lower right):

During this process we want to stay as close as possible to the points in the cloud and apply as
few changes as possible to individually modeled objects while still meeting the rules of the
design software. Another important objective is that the sequence of our modeling actions must
not have any influence on the final result. For example, if we begin the modeling of a long
piperun with a very short cylinder fit along this piperun, the fitting of this initial cylinder axis will
have a small error with respect to the full length of the pipe. Perhaps the pipe is bent or perhaps
strong noise is present in the point cloud. As a result, every additionally connected piece will be
inserted colinear to this first cylinder or at 90 angles with respect to the initial short axis of
the first cylinder extracted. After a few meters or feet we get a noticeable error as seen in the
figure below.

To avoid these troubles means we need an interactive workflow which include global
optimization for an overall best-fit of all objects rather than following the constraints of the
initially modeled piece. A Final result should resemble the model below:
Constraints are applied
one after another
beginning with first
short pipe section.
Point Clouds to Deliverable in AutoCAD Plant 3D
10



An interesting observation: Modeling point clouds for Design software means to find a good
tradeoff between geometrical accuracy and an easy to handle model which looks like the result
of an engineers design process in a green field situation:

Create Plant 3D Objects
Pattern recognition has allowed for extraction of catalog components from the scan data and the
global optimization has allowed for constraints to be applied to the design in order to meet Plant
3D guidelines. The data is ready for conversion to Plant 3D objects.
Generate Standard Plant 3D Objects
Although there are multiple methods for manually routing the Plant 3D objects over the detected
routing line, the kubit team shows a completely integrated approach. PointSense Plant software
will automatically create Plant 3D catalog components from the initially pattern recognized line.
Within the "Export" section of PointSense Plant, the user can simply select "Export as Plant 3D
Objects" and instantly have an intelligent Plant 3D model on screen.
Generate Isometrics
Now that the intelligent Plant 3D model is created, the Autodesk team provides automatic
isometrics and/or orthographic drawing extraction. The "Quick Iso" routine was used to convert
our 3D model into an isometric with bill of materials listed.
Point Clouds to Deliverable in AutoCAD Plant 3D
11


Structural Data Extraction
Structural modeling from point clouds requires very similar tools to piping design.
Pattern recognition algorithms
Defined catalog components (known standards, AISC, etc.)
The kubit software will supply the needed pattern recognition algorithms for recognizing steel
members. The standard AISC steel catalog along with any user defined pattern can be
extracted from the point cloud data.

Fitting requires the user to first define the steel profile desired for search. The user must then
supply two points on the member. This supplies a search region for the software to begin the
search for the desired piece. This process can be faster by performing a batch fit option from
the kubit menu. A standard AutoCAD solid is derive with the proper centerline.
Point Clouds to Deliverable in AutoCAD Plant 3D
12

Locating Tiein Points
If there is one thing we have learned from modeling and/or the intelligent modeling process, it is
that post-processing efforts for generating complete models from cloud data can be very time
consuming depending on the size of the project and level of detail. Often times a designer only
needs to derive specific tie-in points from specific pipe runs in the field. Why model everything in
order to extract a few key points?
Workflow options:
The user needs multiple options for deriving the tie-in points along a run. Based on the scan
position, the user may only be able to see a flange or may only have scan data for the flange
from a specific angle. For this reason, a software must allow flexibility in calculating the internal
tie-in point
.

3 on front of face
3 on back of face
3 on hub
detect cylinder axis +
point on face
detect cylinder axis +
point on back of face
detect cylinder axis +
point on hub
Point Clouds to Deliverable in AutoCAD Plant 3D
13

Design Collaboration and Clash Detection
One of the most important uses of cloud data is comparing existing conditions with proposed
design. In this case, a clash detection tool is extremely valuable. Fortunately, Autodesk provides
this functionality directly in AutodeskNavisworks software. Please read the instructions
below for
performing clash detections between solids/surfaces and a point cloud.

Workflow options:
1. APPEND a DWG to Navisworks which includes both point cloud data and
solids/surfaces of interest
2. Click the CLASH DETECTIVE icon in Navisworks which opens up two columns
3. On the left column, choose the point cloud data referenced in the DWG. Make sure that
the clash setting is marked for POINTS.
4. On the right column, choose the solids/surfaces referenced in the DWG. Make sure that
the clash setting in the column is marked for SURFACES
5. Set an appropriate tolerance for the clash searching between points and solids
6. Set the clash type to CLEARANCE and Run the clash detective
7. Visualize the clashes detected in Navisworks. You will notice the highlighted
solids/surfaces which clash with the point cloud data within the desired tolerance set.