Integration of Robotics & SAUL for Automated Inspection of Composite Parts

Integration of Robotics and Surface-Adaptive Phased-Array UT to Achieve Fully

Automated Inspection of Complex Composite Parts

Michel Brassard, Contour Dynamics Inspection Systems, Levis, Québec, G6V 7M5, Canada and
Techno Diffusion NDT Systems, Montreal, Canada
Deborah Hopkins, BERCLI Corp., Berkeley, CA 94703
Jean-Noël Noiret, EADS Composites Aquitaine, Salaunes, 33160, France

Abstract: Achieving high-speed inspection rates for complex composite parts requires versatile and
integrated systems that meet the challenges of automated part handling and NDT for ship sets that can
include hundreds of different parts. Cost-effective integrated solutions are presented that are being
successfully used to inspect composite parts for fully automated aerospace applications. The most
appropriate solution for a given ship set depends on factors that include detection and sizing
requirements, the range of sizes and geometries of the parts, the required inspection speed and cost
constraints. The challenges of optimizing different technologies and integrating them into a single system
are described for a recently implemented industrial solution. Lessons learned from the project are
presented both in terms of technology integration and implementation of a new ultrasonic software
algorithm. Surface-Adaptive Ultrasound (SAUL) is a very recent advancement in phased-array
technology that is being used to overcome inspection challenges that include highly contoured surfaces;
parts with small radii such as those often found on blades and stiffeners; rough and irregular surfaces
including regions of ply dropoff and lap joints; and parts with varying shape, curvature, and thickness
with length. Although vision systems and robots can be used to achieve highly accurate part following,
the part-to-part variability that is typically encountered with composites creates problems for automated
part and probe positioning, as well as accurate part tracking. This paper demonstrates the performance of
a cost-effective inspection solution for complex-geometry composites in a high-volume production
environment achieved by combining advanced UT technology with industrial robotics and vision
technologies.

Introduction

Aerospace composite manufacturers face the reality that traditional inspection approaches are
inefficient for the new generation of composite parts. The integration of NDE into the
manufacturing process is a potential solution to this problem. Although the ultrasonic inspection
of complex parts using Computer-Aided-Design (CAD) drawings and Teach-and-Learn methods
is frequently used in automated NDT of aerospace components, it is uncommon for these
systems to be required to rapidly inspect a large number of part types with various geometries.
An added difficulty frequently encountered with composites, is that actual part geometries differ
slightly from CAD drawings.

In today's competitive environment, NDT must become part of the design and manufacturing
process to meet requirements for throughput and efficiency. The time required to go from CAD
drawings (which for aerospace purposes are often in the form of CATIA drawings) to inspection
must be reduced. Importation of CAD drawings into NDT systems or teach-and-learn processes
is not sufficient to deal with the increased use of CAD data with greater geometric part
complexities. In the same way that manufacturing was integrated with CAD/CAM systems some
NDT of Composites – Seattle – WA - 13-14 May 2013 Page 1
 Integration of Robotics & SAUL for Automated Inspection of Composite Parts
years ago, the next generation of inspection systems must be fully integrated into the
manufacturing process to meet current production challenges.
Implementation of Processes

Fully integrated Computer-Aided Design and Manufacturing Systems are referred to as

CAD/CAM. CAD involves creating computer models defined by geometrical parameters.
Computer-Aided Manufacturing (CAM) uses geometrical-design data to control automated
machinery. CAM systems are associated with computer numerical control (CNC). Modern
CAD/CAM systems differ from older forms of numerical control (NC) in that geometrical data no
longer has to be encoded mechanically. The analogy to current automated NDT systems is
evident; we are commonly still encoding our systems in isolation without interaction with other
departments.

The system described in this paper is a CAD/CANDT system in which design processes and
non-destructive testing are embedded. Making the analogy to CAM systems, NDT is fully
integrated and human intervention is not required before or during the inspection.
Implementation of a CAD/CANDT system offers a number of benefits including a reduction in
the time required to design scan plans for new parts in addition to a reduced cycle time. Other
integration objectives are sharing information, avoiding duplication of work, reducing wasted
effort, eliminating non-value activities, and standardizing software suites, which as a
consequence frees up valuable NDT resources to perform NDT evaluations.

The processes required to perform an inspection are schematically shown in Figure 1. These
processes are no different than what is done for a part to be machined.

NDT of Composites – Seattle – WA - 13-14 May 2013 Page 2

 Integration of Robotics & SAUL for Automated Inspection of Composite Parts

Figure 1 – Process Schematic

The first step is to import a CAD/CATIA drawing into the path-generating software (Mastercam)
to create a tool path or from an NDT perspective, a scan plan. The scan plan is converted into
CNC codes to inspect the part. Conversion is done using a post processor that translates tool-
path data from Mastercam into CNC codes that a controller such as the Siemens SInumerik can
interpret (ISO-programming codes or G codes). These codes are what the operator sees when
the machine is inspecting. ISO codes include instructions on where to move, how fast to move,
and what path to follow. In essence, the system becomes a machine tool. We therefore
embrace machining technology without performing manufacturing.

For example, Figure 2 shows a practical implementation of a system that is using the above
technology to automatically plan, load, inspect and unload composite parts. This is what is
loosely called a CAD/CANDT system.

NDT of Composites – Seattle – WA - 13-14 May 2013 Page 3

 Integration of Robotics & SAUL for Automated Inspection of Composite Parts

Figure 2 - CAD/CANDT System

Integration Challenges

Integrating multiple technologies into a single system is a complex and challenging operation.
Learning how to make effective use of existing CAD/CAM technologies and adapting them to
NDT requires some planning, as processes are different. A machining program, by definition,
does everything from the front surface. However, in NDT we must control what goes inside the
material. For example, when inspecting a radius the coverage of the back surface is quite
different than what is seen from the front surface. We must therefore take into consideration the
ultrasonic physical phenomena when planning our scan plan. The challenges that we face
include:

I. Limited familiarity with multiple disciplines. CAD/CAM technicians and engineers are
not usually proficient in NDT. Similarly, NDT personnel are not typically familiar with CAD
intricacies. This is a major stumbling block in technology integration. It can be difficult to
exchange information and communicate when each party does not understand the other
party’s field of competence.
II. Software. NDT is a niche sector, and not the primary market for standard CAD/CAM
software packages. Although these packages typically have inspection tools, they were not
part of the developer’s original focus, making NDT customization more difficult. It is,
however, still doable.
III. Time. To successfully incorporate beneficial technology requires an investment of time
up front during pre-production planning. The payoff is that the NDT system cycle is much
faster because the machine utilization is fully dedicated to inspection.
IV. Training. NDT personnel are inexperienced in the use of machine codes. In the same
way that machinists have learned to interpret and use these tools to their benefit, NDT
personnel must be trained to use these tools effectively for inspection.

NDT of Composites – Seattle – WA - 13-14 May 2013 Page 4

 Integration of Robotics & SAUL for Automated Inspection of Composite Parts

CAD/CANDT has the potential to eliminate some of the inefficiencies that arise from the way
NDT is currently being implemented into production. Some questions we in the NDT community
must ask ourselves:

• Is NDT being considered as part the design process to facilitate inspectability?

• Is the inspection machine time used efficiently?
• Why is it that the tool paths for machining are done by the CAM office whereas
inspection tool paths (scan plans) are done by NDT personnel?
• Why are we using the inspection system to simulate inspection rather than using
commercially available software commonly used for machine tools?

Functionally, for integration we need to look closely at the points of failures/weaknesses of the
system. Technologies can work together, to the benefit of all, if CAD, CAM and NDT personnel
are working in closer cooperation. To perform effectively, a manufacturer needs its teams to
operate as a unified group; working towards common business goals, sharing information about
the tools that are used. NDT cannot and should not operate in isolation.

Ultrasonic Inspection

Composite parts inspected using the immersion system represented in Figure 2 have complex
3D shapes and include tight radii. Inspection of these parts using a conventional phased-array
approach requires a combination of curved and flat probes. The approach is illustrated in the top
row of Figure 3 for a hat-section stringer where linear and curved arrays have been optimized
for each radius and the flat sections (as shown in the schematic diagram). Accurate positioning
of the curved arrays is essential for obtaining accurate results, and this can be particularly
difficult on long composite parts. When accurate positioning is maintained, very good results
can be obtained. The time-of-flight and amplitude C-scans shown on the right are from a hat-
section test specimen with known defects, all of which were successfully detected and imaged
using the probes illustrated in the schematic drawing. The middle row of Figure 2 shows B-
scans for the top convex radius obtained for an actual composite stringer using a linear probe
(as illustrated in the left-hand photograph). B-scans were obtained with the linear array without
SAUL (middle image) and with SAUL (right-hand image). Applying the surface-adaptive SAUL
algorithm (described later in the paper) allows a backwall signal to be measured, but the lateral
extent over which there is a strong backwall signal is relatively short. The uniformity of
measured signals and the lateral extent of the backwall signal are greatly improved using a
4x16 matrix array optimized for use on aerospace composites (a schematic drawing of the
matrix probe is shown in the bottom-left corner of the figure).
Three Phased-Array Solutions for Inspection of a Hat-Section Stringer

NDT of Composites – Seattle – WA - 13-14 May 2013 Page 5

 Integration of Robotics & SAUL for Automated Inspection of Composite Parts

Schematic drawing of linear and curved arrays used to inspect a composite stringer (without SAUL) and
the resulting scans where all embedded defects were detected and imaged

Top convex radius inspected with a linear probe without (middle) and with SAUL (left-hand image)

Convex (middle image) and concave radii (left-hand image) inspected with SAUL and a 4x16 matrix array
(lower-left figure)

Figure 3 – Phased-Array Solutions for Inspection of a Hat-Section Stringer

Surface-Adaptive ULtrasound (SAUL)

The innovative SAUL technique developed by the CEA and implemented in M2M
instrumentation is used as a solution to the problems posed by complex geometries including
probe positioning and part variability [2]. The objective of the adaptive technique is to generate a
wave front that is normal to the front surface of the test specimen. The specimen shape is
estimated in real time from the front-surface echoes. An iterative algorithm then optimizes delay
laws based on minimizing the error function determined from calculating the travel times to the
surface, which is also done on the fly. The delay laws are continually updated to adapt to the
changing geometry parameters [3]. The increase in processing time is partially offset by a
reduction in measurement time (compared to electronic scanning) achieved by firing all
elements at once rather than firing by groups. SAUL has been automated in the M2M

NDT of Composites – Seattle – WA - 13-14 May 2013 Page 6

 Integration of Robotics & SAUL for Automated Inspection of Composite Parts
instrumentation so that the user only needs to input a few parameters as part of the ultrasonic
set-up process.

SAUL’s Ability to Compensate for Probe Misalignment

As discussed above and reported by Hopkins et al. [1], probe positioning is critical for obtaining
satisfactory results particularly for curved arrays. The complex shapes of composite parts and
part-to-part variability increase the positioning challenge especially for long parts. The ability to
fully automate inspection processes therefore depends on being able to address positioning
errors and part variability. One automation approach is to install extremely precise positioning
systems. Such systems work well, but do not address part-to-part variability. It is also common
for systems to have features that allow repositioning of the part or adapting the scan plan where
necessary. This, however, requires time and actions that must be performed in the tank, leading
to an increased cycle time and loss of availability of the system to inspect parts.

Figures 4 and 6 show the results of experiments performed to demonstrate SAUL’s ability to
correct for probe misalignment. The top figures in Figure 4 show the position of a linear probe
for two measurements, as well as the delay laws applied to the probe’s elements (indicated by
the red bars). For the first measurement the probe was oriented parallel to the composite-plate
test specimen and fired with no delay laws applied. The resulting B-scan is shown in the lower-
left image. For the second measurement, the probe was tilted 5 degrees from normal as
indicated in the top-right image. In this case, SAUL was activated to calculate the delay laws
required to compensate for the tilt of the probe. As described above, the objective of the SAUL
delay-law optimization is to generate a wave front that is normal to the front surface of the test
specimen. The same principle is used for curved surfaces, where the incident wave fronts are
optimized locally to match the surface (see Figures 5 and 6).

NDT of Composites – Seattle – WA - 13-14 May 2013 Page 7

 Integration of Robotics & SAUL for Automated Inspection of Composite Parts

Figure 4 - SAUL Correction for a Misaligned Linear Array

The experimental setup for measurements made with a curved array and a composite tube are
shown in Figure 4 (top images), along with a picture of the release-film defects that were placed
mid laminate in the tube (bottom image). The probe curvature is a good match to the tube, and
all four defects were detected and imaged when the curved array was well centered on the test
specimen as shown in the first column of Figure 5. When the probe was shifted 3 mm off
center, all return signals were lost as shown in the center column. Maintaining the probe in the
offset position and activating SAUL completely restores the signals as shown in the third
column. Comparing the images in the first and third columns shows that the scans obtained with
SAUL for the probe in the offset position agree very well with the images obtained when the
probe was correctly centered on the test specimen.

Figure 5 - Experimental Setup and Defects for Scans Shown in Figure 5

B-scan: B-scan: B-scan with SAUL

Curved array well-centered Curved array offset 3 mm Curved array offset 3 mm

Time-of-Flight C-scan Time-of-Flight C-scan Time-of-Flight C-scan

NDT of Composites – Seattle – WA - 13-14 May 2013 Page 8

 Integration of Robotics & SAUL for Automated Inspection of Composite Parts

Amplitude C-scan Amplitude C-scan Amplitude C-scan

Figure 6 -. SAUL Compensation for a Misaligned Curved Array

NDT of Composites – Seattle – WA - 13-14 May 2013 Page 9

 Integration of Robotics & SAUL for Automated Inspection of Composite Parts

Figure 7 – 3D and Top Views of an L-Shaped Composite Part with a Tight Radius

Amplitude C-Scan Time-of-Flight C-Scan

Figure 8 – Amplitude and Time-of-Flight C-Scans for the Part Pictured in Figure 7 Obtained with
a 2D Matrix Probe and SAUL

Amplitude C-Scan Time-of-Flight C-Scan

NDT of Composites – Seattle – WA - 13-14 May 2013 Page 10

 Integration of Robotics & SAUL for Automated Inspection of Composite Parts
Figure 9 - Amplitude and Time-of-Flight C-Scans Obtained with the Matrix Probe and SAUL for
the Same Part Shown in Figure 8 for the Part Position Shifted 5mm to Simulate
Misalignment of the Probe

The results of laboratory experiments and modeling were used to optimize SAUL for the wide
range of composite-part geometries encountered in aerospace. This effort included M2M’s
design and optimization of the 4x16 matrix probe shown in the lower-left corner of Figure 3. This
probe is being used to inspect hundreds of different composite parts in the first fully automated
SAUL implementation at the EADS composites production facility in Aquitaine, France. The
advantage of the 2D array is that geometry compensation is performed in two directions,
allowing the probe to be used on a variety of complex parts including those with small radii. The
probe also successfully compensates for part variability, greatly improving productivity
compared to installations using multiple probes and/or manual part manipulation.

Figure 7 shows two views of a part inspected using SAUL with the 2D matrix probe described in
Figure 3. The part is a composite L-shaped panel with a tight radius, most easily seen in the 3D
image (left-hand picture). Although not easy to see in the figure, the part is curved in the X-Y
plane. Amplitude and time-of-flight C-scans obtained for the part are displayed in Figure 8. For
ease of interpretation, the C-scan images have been separated into three sections: the top
section corresponds to the vertical wall (short leg of the “L”), the middle section corresponds to
the radius, and the bottom section corresponds to the large panel region. Without SAUL, it was
difficult to obtain acceptable results for the radius section. With SAUL, full coverage of the
radius is achieved as can be seen in the C-scans.

To simulate probe misalignment arising from part variability or poor positioning, the part was
shifted by 5 mm in the long direction (X axis) and re-inspected with the probe position and scan
plan unchanged. The shift in position results in the probe being at the wrong angle with respect
to the part because of the curvature in the X-Y plane. This is evident in the time-of-flight (TOF)
C-scan (right-hand image in Figure 9), particularly for the vertical wall and radius sections. The
non-uniform time of flight data indicates that the relative position of the probe and part has
changed and is different from the scan plan (compare to the TOF C-scan shown in Figure 8
obtained with the probe correctly aligned). In spite of the misalignment of the probe, the
amplitude C-scan obtained using SAUL (left-hand image in Figure 9) compares very well to the
C-scan obtained with the probe correctly positioned (left-hand image in Figure 8). The
agreement of the two amplitude scans demonstrates that SAUL has successfully compensated
for the misaligned probe.

Conclusions

Advanced technologies don't necessarily equate to a state-of-the-art system. Integration is the

key. Not only must the NDT solution be capable of meeting the inspection challenges, but the
integrated system must also be able to meet the production requirements. Quality must be
ensured while also maintaining high production volumes. CAD/CANDT reduces the entire
manufacturing cycle time, as NDT and CAD personnel become part of a unified team. SAUL
has the ability to compensate for misalignment and part variability, and is therefore an important
tool in achieving system compliance in the face of industrial realities. A fully integrated
CAD/CANNDT system with SAUL is an industrial solution that has been demonstrated to be
capable of meeting today’s demanding challenges of composite manufacturing.

NDT of Composites – Seattle – WA - 13-14 May 2013 Page 11

 Integration of Robotics & SAUL for Automated Inspection of Composite Parts

NDT of Composites – Seattle – WA - 13-14 May 2013 Page 12

 Integration of Robotics & SAUL for Automated Inspection of Composite Parts

References

[1] D. Hopkins, G. Neau, W. V. Johnson, L. Le Ber, and P. Calmon, “Surface Adaptive

Ultrasound for Phased-Array Inspection of Complex Composite Specimens,” presented at
Review of Progress in Quantitative Nondestructive Evaluation, Burlington, Vermont, 2011.

[2] D. Hopkins, G. Neau, and L. Le Ber, “Advanced Phased Array Technologies for
Ultrasonic Inspection of Complex Composite Parts,” in Proc. of International Workshop on
Smart Materials, Structures & NDT in Aerospace, NDT in Canada 2011, 2-4 Nov., Montreal,
Quebéc.

[3] S. Mahaut, O. Roy, S. Chatillon, and P. Calmon, “Modeling and Application of Phased
Array Techniques Dedicated to Complex Geometry Inspection,” in Review of Progress in
Quantitative Nondestructive Evaluation 21, edited by D. O. Thompson and D. E. Chimenti, AIP
Conference Proceedings vol. 615, American Institute of Physics, Melville, NY, 2002.

NDT of Composites – Seattle – WA - 13-14 May 2013 Page 13