Robotic planning of inspection paths is often a tedious task. Most robotic systems are manually programmed or programmed using a teach and learn approach where inspection points are input manually. For parts with small radii or with complex curves, it is impractical to rely on a point-to-point teaching. Some general computer aided manufacturing software exist and are sometimes used for this purpose but they are designed to be operated by a CAM specialist and do not take into account the ultrasonic specificities.

For effectiveness, the UT path programming process must achieve full automation with minimal intervention, use the part geometry as the guide, fully control the part with minimum number of passes and be user-friendly for a quality technician or engineer.

First Task - Transformation in 2D space

Optimizing machine utilization is critical in today’s manufacturing. Setting up the scan plan off-line using the 3D part surface drawing is the best way of reducing non-productive machine time. To implement such a mathematical simulation, the inspection parts are decomposed by meshing the surface. As all computations are based on the mesh quality, it is important that they be formed of quasi-equilateral triangles.

Inspection analysis remains easier in 2D space and, furthermore, all ultrasonic software packages are tailored to a 2D environment. As a robotic system works in 3D space, a streamlined conversion of 3D to 2D data is pivotal. The solution is based on a method, common in 3D animations, where conformal map projections are used to preserve local mesh angles and minimize distortion. Even though the overall shape changed, all grid lines still intersect at 90°. The 2D flattening preserves overall shapes to keep the flattened surface recognizable by the inspector. Not only does it make easier to validate the scanning path, but it also facilitates locating indications and decision-making when the final ultrasonic data is finally represented on a C scan. This representation, which is common in ultrasonic testing, is defined by the specification BS-EN 1330-4:2000 as the image of the results of an ultrasonic examination showing a cross section of the test object parallel to the surface. The flattening file is then saved as a textured OBJ file where each vortex has both 3D and 2D coordinates.

Path Generation

Paths are generated automatically based on different algorithms selected by the user. Unfortunately, it is not possible to have one method fits all for all parts due to the variety of shapes that must be inspected. The process begins by defining the ultrasonic parameters such as probe diameter, delay law configuration, frequency, water column, overlap, etc. The path indexation is calculated from a simple beam model to find the most off-center aperture beam that still receives an echo on at least half of its elements. Combining this with an approximation of the beam width, the covered part width is estimated. This is repeated for all positions along a scan line to lead to the minimum covered width as the indexation value. This technique allows you to generate a scan path that adapts to the part’s geometry and minimizes the number of passes

For example, the inspection of a cylinder that has a varying diameter leads to an inspection path that requires curvature following in two directions. The parameters used by the algorithm includes part curvature, normal vectors, principal direction and edges. Most parts can be mathematically analyzed using these elements. Usage may lead to requirements for additional elements to be integrated.

The parameters are applied based on the selected path algorithm and results in a proposed scanning plan. A brief description of the methods follows:

1. Raster method. The 2D raster method aims at reproducing the raster scan path that is traditionally used in ultrasonic testing. The path is constructed in 2D where the longest direction is identified using principal component analysis and used as the scan direction. The perpendicular direction is then defined as the index direction. The user may inverse the direction and keep the distance between the scan line constant or adapted to the curvature.
2. Minimum curvature 2D or 3D. For these two methods, principal curvatures and principal directions are used. The principle is to create scan lines that stay as flat as possible. This implies that the scan lines follow the direction of least curvature. Using the principal direction field, this is accomplished by moving along the surface in small increments and adjusting the scan direction so that it points toward the principal direction. These methods generate scan lines which are no longer straight (neither in 2D nor in 3D), but follow the surface more closely.
3. Sweep method. The initial scan line is calculated using the principal direction of least curvature on the 2D mesh. This line is swept across to fully cover the part.
4. Edge Following. This method is more appropriate for long and skinny parts and for parts with non-parallel radii. It uses the longest side as the starting-point boundary. The side is swept across the part using curvature rules.

Not all parts can be done as one part. One option is to cut the part in sections and to stick them together. Mathematically, a cluster analysis is done to break down the part in sections. In this example, the part is broken down in five regions as shown by the color on the picture. When the part is broken down into sections, the algorithms work separately in each region.

Planning concept

The concept may be summarized as:

1. Start with the part 3D CAD.
2. Extract the surface CAD. This is the surface where the sound beam penetrates the part.
3. Mesh the part.
4. Flatten the part to allow 3D to 2D transformation. Minimize distortion.
5. Set the ultrasonic parameters.
6. Select the algorithm best adapted to this part.

Scanning

The path must be converted to software codes that can be read by a motion controller. Post-processors are used to translate the motion commands into a specific program tailored to a robot or a controller type.

The part must be positioned in a simulation software based on the robot (or the Cartesian) workspace and its end effector. The calibrated robot tool reference (Tool Center Point or TCP) allows the user to verify that the part can be inspected using the generated scan plan. Minor modifications are done to verify mechanical interferences, joint limits, and singularities. For ultrasonic scanning the TCP continually moves along a defined path, therefore singularities may happen as the robot cannot be in joint mode (a mode used to go from Point A to Point B without control of the path to reach these positions).

Once the path is corrected and verified, the part is positioned in the workspace. The part position, if not assured by a precise jig, may be controlled with a touch probe or better with the ultrasonic transducer. In this last option, the part’s position is detected through a quick ultrasonic scan that allows you to collect partial, but sufficient data to determine its misalignment relative to an ideal positioning, and to modify the inspection path program accordingly.

Conclusion

Ultrasonic robotic path planning may be complex, especially for pulse echo ultrasonic inspection where probe to part perpendicularity is crucial. For ultrasonic scanning, the continuous path has challenges due to complex part shapes and new materials, in addition to mechanical constraint and robot singularities. Machine operating time optimization is an essential objective and offline ultrasonic scan plan generation is an integral part of the process. The presented software approach with the latest advances in robot planning meets these criteria.