Originating Author: Frank Maslar August 17, 2016 Collaborative Robots & Automation Technologies Automotive Manufacturing 2016 Vision Based Error Proofing Techniques Frank Maslar Vision Systems Technical Specialist Ford Motor Company Powertrain Manufacturing Advanced Manufacturing Engineering Email: [email protected]
19
Embed
Automotive Manufacturing 2016 Vision Based Error Proofing ...
This document is posted to help you gain knowledge. Please leave a comment to let me know what you think about it! Share it to your friends and learn new things together.
• Powertrain: Engine, transmission and drive axle components • Repetitive – similar parts are always processed
– Lines can handle part type variations – Generally metal parts with differences in surface finish
• Location variation less than 5mm in all directions • Parts stay in station 20-30 seconds
– Typical vision processing time can be up to 5 seconds
Originating Author: Frank Maslar August 17, 2016
4
Automotive Powertrain Manufacturing Environment
• Lines run 24/7 – Ease of recovery from “vision failures” is a requirement – Downtime is expensive
• Vision is included in the standard (required) process – Bypass buttons have been removed – 6 σ performance ( 3 defects/million) performance is required
Originating Author: Frank Maslar August 17, 2016
5
Vision Applications in Assembly
Average 100 industrial cameras in each line • Standard system is PC based using 5MP GigE cameras • 50 cameras read Data Matrix codes
– Standard data matrix code 20 rows X 20 columns – Used to uniquely identify each component
• Complete part traceability and assembly birth history • Make sure the right part is used in the assembly
• 50 cameras verify proper assembly – Parts fully assembled – Manage complexity using color if parts cannot be marked with the data matrix
Future Technology Opportunity 3D Data Visualization
Most 3D cameras represent 3D data as traditional 2D images where distance is converted to gray scale • 3D data is displayed in 2D • 2D image processing tools are used to analyze the data • Issues:
– Some 3D details are lost in the conversion – Traditional 2D image processing tools do not take advantage of the true
3D data
Originating Author: Frank Maslar August 17, 2016
16
Future Technology Opportunity 3D Data Visualization
The need is to develop systems that work in 3D space • Data should be visualized as a solid model • Visualization tools should be able to manipulate the data like a solid model
including zoom, pan and rotate • 3D image processing tools need to be developed to analyze the 3D data
– Edge tools and 3D distance measuring tools – Blob tools which characterize the object in 3D – 3D Pattern matching tools
Originating Author: Frank Maslar August 17, 2016
17
Future Technology Opportunity Vision Data Sensor Fusion
Vision data can come from a variety of sources • Traditional 2D gray scale images • 2D color images • 3D information
– Multi-camera or stereo – 3D point cloud
• Laser type scanner • Area camera with structured light
• Each data type is typically processed in its native environment using tools developed for this environment
Originating Author: Frank Maslar August 17, 2016
18
Future Technology Opportunity Vision Data Sensor Fusion
The need is to combine these data types into a unified workspace • New image analysis tools are required to analyze the data in this unified
workspace • Tool Requirements:
– Intuitively easy to use – Seamlessly work with all the data types