Developing the WiseWELDING Machine Vision System for Adaptive Robotic Welding
Print
Author(s):
Robert Modic, M.Sc. -
Wise Technologies Ltd.
Industry:
Machine Vision/Imaging, Manufacturing
Products:
NI 9401, LabVIEW, PCI-7811R, Vision Development Module, FPGA Module, NI 9426, cRIO-9151
The Challenge:
Replacing manual operations with an automatic welding path correction used in high-quality stainless process equipment production while designing, prototyping, testing, and deploying a complete system to production in only two months.
The Solution:
Developing WiseWELDING, a 3D machine vision system that adapts robotic welding paths to geometry variations between consecutive parts, compensates for large geometry displacements within the working window (50 by 40 mm), and senses gapless "butt" joints (gaps greater than or equal to 0,05 mm) all in one.
"WiseWELDING was able to comply with all the specific customer requirements and achieve 10 times throughput gains compared to manual adaptation in only two months."
About Robotic Welding
Robotic welding introduces many advantages to the manufacturing process: smooth movement, speed, precision, repeatability, flexibility, and resistance to hazardous environments; yet, the paramount prerequisite for any successful advanced application is mastering welding technology. The main motivation for the solution was the fact that part geometries in welding production commonly vary to some extent. Consequently, actual robotic welding paths have to be corrected for each specific part. Correction can be conducted manually for each path point using a robotic tool tip and subjective judgment of the operator. This approach is time-consuming, repetitive, and prone to human error.
Automatic welding path correction can be achieved using add-on vision modules for noncontact sensing of actual path points, which makes correction more precise, repeatable, and usually 10 times faster when compared to manual operation.
Considerations for Choosing the Development Platform
We used the NI development platform because it offers high-level rapid application development programming, a flexible hardware platform, and tight hardware and software integration.
In addition, the NI PCI-7811R R Series multifunction data acquisition board with the NI 9151 R Series expansion chassis for NI C Series I/O modules offered the most flexible and versatile I/O interfacing with the robot. We developed the application using the NI LabVIEW graphical programming environment, the NI Vision Development Module for image processing, and the LabVIEW FPGA Module for custom I/O. Using the third-party gigabit Ethernet driver support for LabVIEW, we easily integrated the high-end vision sensor used for image acquisition.
WiseWELDING
The first step to adaptive welding using WiseWELDING is a seamless robotic platform upgrade with a vision system followed by teaching manufactured part's master geometry using machine vision sensing. Teaching takes place only once and is generally much faster due to its noncontact nature. After completing the series, welding can take place and the system will automatically compensate for geometric variations between consecutive parts.
All standard seam topologies are covered: butt, V-shaped butt, overlapping, and corner joint, and custom seams and their various start/stop configurations can be adapted as well. NI high-level rapid development tools are invaluable in this segment. Whether it is robotic communication, custom triggering and I/O for achieving real-time performance, or designing and prototyping custom vision and data processing algorithms, we can deploy, test-drive, optimize, finalize, and implement functionality in hours. The LabVIEW Instrument I/O Assistant, field-programmable gate array (FPGA) execution on a simulated target, the NI Vision Assistant, and signal processing libraries quickly yielded usable code, which saved valuable time that we would have spent coding for the desired functionality. This groundwork and using NI advanced vision functions such as an "advanced edge finder" are further supported by custom-developed image and signal processing functions to properly correlate multimodal data obtained by our imaging module, segment, and register features, and allow for powerful real-time performance in varying conditions, from black metal to stainless steel, mate or brushed, or with mixed surface reflectance regardless of scratches, irregularities, and ambient conditions.
Conclusion
The current system runs on a multicore, PC-based processing unit with a Windows OS, but due to timing, reliability, and form factor constrains, we are considering using an NI real-time OS-based system platform, such as the NI EVS-1464RT, for our next implementation. With the chosen NI hardware and software development platforms, we can easily envision further development, optimization, and customization of the current product.
Related Case Studies
Developing a High-Performance Falling Objects Sensing System Based on Machine VisionNI Vision and LabVIEW Extend the Limits of Machine Vision Accuracy
NI Compact Vision System Helps Achieve Six-Sigma Repeatability Standards in Inspection of Automobile Spark Plugs
Neutrinos, Robotics, and Machine Vision
Building an Educational Machine Vision Tool Using LabVIEW
Print
|
|