>
Geometric Inspection of small components with �CT-scanners Solving leakage problems on
car door assemblies
Manufacturers know when something goes wrong. Finding out why is often a much
more involved process. The ensuing course of investigation consumes time, money, and
resources. Even once the problem has finally been identified, making quantified
adjustments is yet another stage en route to rectification. The manufactured part is
different from the designed part, or different to how it should be or has been in the
past, but by how much, where, and in what direction? For a major Australian car
manufacturer, identifying the cause of leakage on door assemblies was a priority.
PolyWorks provided the answers in less than 1 hour.

About Scanning and Inspection
Pty Ltd.

Headquartered in Melbourne,
Australia, Scanning and
Inspection Pty Ltd distributes
3D measurement equipment
such as portable CMM arms,
laser trackers, and laser
scanners. Scanning and
Inspection is the master
distributor of PolyWorks in
Australia.

www.s-i.com.au
Data acquisition

To quickly and efficiently capture the surface information,
specialists at Scanning and Inspection used a laser scanner
mounted on an articulated arm. The type of laser scanner
utilized for this application was the ModelMaker Z70, a high-
quality scanner capable of digitizing a surface at a
remarkable density of 0.05mm x 0.05mm, which is around
400 points per square millimeter. This results in a realistic,
full, 3D digital representation of the part, a plethora of data
just waiting to be transformed into useful information. The
scanner gets its 3D reference from a portable CMM, in this
instance a Platinum FaroArm, which measures to a
repeatability ranging from �0.0051mm depending on its
size. The FaroArm and MMZ70 Laser Scanner were wheeled
into position onsite thanks to a heavy-duty tripod, hence
the hardware could be conveniently set up adjacent to the
car doors in question (Fig. 2). Capturing the required
surface data took only a few minutes per car.

Description of the investigation
An automobile manufacturer recently experienced one of
those unavoidable manufacturing scenarios. They were
encountering variations in the alignment of certain
passenger cabin door seals that they were producing and
assembling. It was apparent that certain cars that had
been manufactured during one specific facet of the car
production were as expected, the doors and seals all lined
up with the rest of the car body seamlessly. However,
those coming off the production line at a later stage were
experiencing water leaks during the testing stage. The first
task concerned locating where the water was getting
through, but the more involved analysis was in regards to
the relationship between the doorframe and the car body.
Which of the two was different to that of the vehicles that
were not leaking, by how much, where, and in what
direction?
If one batch of the automotive assemblies were coming out of production and were
experiencing no leakage problems, and others being produced at a later stage were leaking,
then they must somehow differ in their structure (Fig. 1). How does the problematic door
frame assembly differ from its good car counterpart? In order to learn what changes need to
be made during production, the client needs to be able to quantify the discrepancy. The
relationship and the variations between the good and bad door frame assemblies need to be
evaluated. A detailed comparative analysis was acquired, and thanks to the latest in laser
scanning technology and the PolyWorks suite, it was achieved within hours.

Fig. 1: Car assembly
Fig. 2: ModelMaker Z70
mounted on a FaroArm The PolyWorks Process

Once the scan data had been collected, the post-processing stage of the analysis began.
Using PolyWorks, specialists at Scanning & Inspection virtually aligned the laser scanned
data from the leaking car to the one of the good car, and performed advanced
measurement and comparison analyses. However, prior to performing the comparative
analyses, a series of point cloud manipulations had to be conducted in PolyWorks.

Refining the scan alignment

During the data acquisition stage of the job, several scan passes were obtained. Each scan
patch is automatically aligned using the articulated arms native alignment technology that
provides the spatial reference for the changing location of the laser scanner (Fig. 4). Most of
the point clouds acquired by the laser scanner sit on top of others that have already been
accounted for by other passes of the scanner. Certain portions of the part are consequently
measured more than once. These areas of multi-tiered data are referred to as areas of
overlap.

To refine the aligned point cloud obtained from the portable CMM, the PolyWorks IMAlign
module uses these areas of overlaps to realign all scans. First, a deviation analysis was
performed on the aligned point cloud. PolyWorks automatically computes the average
distances between each overlapping scan and displays a color that represents a deviation
value. In a glance, users can identify the scans that are out of tolerance and that can be
improved through realignment. (Fig.5)

Then, using the PolyWorks best-fit alignment technology, a re-alignment was performed.
To prevent sliding between patches in the same neighborhood, eight control points are
automatically set for each scan, which constrain the degrees of freedom. This feature
differentiates PolyWorks from other scan alignment solutions, which often encounter sliding
problems during scan alignment. After the constrained best-fit alignment is performed, a
new deviation analysis is computed and a color map is displayed. Figure 6 shows that all
scans were perfectly realigned to each other and are now within the tolerance value.

The last step consists in creating a single-layered point cloud model by intelligently
averaging all over-lapping scans using PolyWorks reduce overlap technology. The resulting
point cloud can now be transformed in a surface that will be used for further analysis such as
door-to-door comparisons.
The scanning methodology was the same for both the
leaking and the good cars. All peripheral trim was
removed from the windowframes and doorframes to
expose only the post-production structural steel
components. The doors were closed throughout the data
capture process and the relevant areas associated with the
leaking seals were scanned using the MMZ70 laser (Fig. 3).
As the laser stripe is moved across the surface, collecting
tens of thousands of 3D points per second, the subsequent
surface immediately appeared on the computer screen.

The capabilities of this real-time point cloud surface display
means that any areas that were missed, or still needed to
be scanned, could be identified at the time of
measurement and consequently accounted for.
Fig. 3: Data acquisition Fig. 4: Point cloud data
obtained from portable CMM
Fig. 5: Deviation analysis
showing scan alignment
Fig. 6: Deviation analysis
after best-fit alignment
Creating a polygonal model

In order to compare the surfaces of the good and bad doorframe assemblies, Scanning and
Inspection needed to convert the point clouds to something it could use as a surface. PolyWorks
converted the tens of millions of points from the point cloud into a triangulated mesh in the blink
of an eye. The software used each of the points as nodes for the millions of tiny triangles it
constructed (Fig. 7).

To accurately describe the various shapes of the scanned object, PolyWorks uses fewer triangles
for the more consistent sections of the cars surface and more triangles when the contouring
gets more complex, all based on tolerance values specified by the user. Nearest neighbor and
search distance parameters are also stipulated prior to the conversion and to create the
triangulated surface and to reduce the inherent digitizer noise to a maximum.
Door-to-door comparisons

Using a complex series of least squares mathematics and surface analysis the two newly
triangulated and optimized laser scan datasets were laid one atop the other, and aligned using
the PolyWorks best-fit alignment technology. PolyWorks then developed a color-coded
distribution pattern highlighting the comparison of the two assemblies. The re