Today’s manufacturing environment is fast paced and filled with various forms of technology to help manufacturers monitor and control their processes. Machine vision has become one of the mainstays for process control, metrology and assembly verification in today’s modern manufacturing facilities. Higher adoption levels of machine vision hardware and software is a positive trend for professionals that believe in the benefits of machine vision for non-contact inspection and metrology.
However, when technology is rapidly adopted there are opportunities for technologies to be mis-applied or stretched beyond their current capabilities. One classic example of this “capability stretch” is the application of two-dimensional (2-D) machine vision solutions for complex parts and processes that truly require three-dimensional (3-D) image processing and algorithms. An area where this issue is best evidenced is in assembly verification applications.
The definition of assembly verification is to inspect parts as they are put together to make sure the necessary components are present and/or “assembled” correctly. Most assembly verification is done in line or near line using three methods: human inspectors, mechanical probes and 2-D vision. For in line assembly verification, a “go/no go” determination is made before the assembly is released to the next operation. This article will focus primarily on in-line assembly verification using machine vision.
Parts that are assembled today are becoming more and more complex, requiring more assembly verification. The more complex parts also have a large degree of three-dimensional features that, if captured in 3-D, can be used to help make the overall inspection of these assemblies much more robust. If the assemblies are being processed with mechanical probes or 2-D tools alone, manufacturers are missing out on a significant amount of 3-D data that can be used to aid in process troubleshooting and error proofing.
Using only 2-D data also can result in a higher degree of false positive line stoppages and unnecessary down time due to failures of the 2-D vision systems caused by poor image quality and inconsistent component contrast. Vision systems that can generate and use 3-D data for assembly verification will be more robust, simpler to use and lower cost-to-own over time.
The Case for 3-D Machine Vision in Assembly VerificationIt is impossible to ignore the indelible mark that 2-D machine vision has made on manufacturing. No one can argue against the fact that properly applied 2-D machine vision is an enabler to higher throughput, lower cost of manufacturing and improved overall quality. In the past 10 to 15 years there have been tremendous advances in camera resolution, processing power and algorithms.
However, 2-D machine vision systems still suffer from many of the traditional short comings that lead to customers having less than stellar implementation and ownership experiences. For example, minor changes in part appearance or part presentation can lead to large changes in measurement reliability and robustness. These “appearance” issues can often be deadly to 2-D systems that rely on contrast to make their decisions. When 2-D machine vision is applied to a complex 3-D problem, many of the common 2-D issues become even more evident and often stretch the capability of the 2-D systems.
This stretch of 2-D technology can lead to a sub-optimal system as the customer is required to add more cameras and expensive lighting to the station until it resembles an extremely complicated and costly discothèque. The end result is a system that costs significantly more and is too temperamental and complex to meet the technical requirements of the manufacturer. It is these applications where 3-D assembly verification is truly a better place to start. Where traditional 2-D vision uses carefully orchestrated external lighting to create contrast between specific part features, 3-D systems can use integrated lighting to create multiple and distinct patterns resulting in a high immunity to changes in both part appearance and presentation.
Where traditional 2-D machine vision systems generally make inspection decisions based on a single image, 3-D systems can use dozens of images that can be layered to build topography up from the assembly itself. Where traditional 2-D machine vision provides an operator interface dealing in grey levels, pixels, and thresholds, 3-D systems can offer intuitive answers based on 3-D models of the part built up from the part geometry itself. These 3-D benefits are particularly valuable when vision is used to inspect complex assemblies, where the assembled part often results in several distinct shapes.
These shapes create a topography that is recognized as unique when processed with 3-D vision systems. The part topography can be said to create “3-D relief” that can be collected in 3-D and used for multiple processing of the complex geometry. This 3-D relief can be used as additional data in assembly verification, allowing the vision system to process the same part several different ways for the most robust solution possible. This multi-processing of the image data also helps make sure that the 3-D machine vision system will not get “fooled” by some of the more prevalent issues that plague 2-D machine vision, making 2-D systems earn the moniker of “temperamental 2-D” in some markets.
Temperamental 2-D in Automotive PowertrainToday, 2-D machine vision is widely used in automotive powertrain manufacturing with a high degree of success. In some cases there are as many as 80 to 120 cameras in a powertrain assembly plant doing applications like bar code scanning and presence/absence of major components. Many of these applications are perfectly suited for today’s powerful 2-D systems; however some assembly verification applications are not in the sweet spot for 2-D vision.
In general, 2-D machine visions systems attempt to solve assembly verification problems using a combination feature based measurements like edges or corners. These feature based measurements count on the 2-D vision system being able to reliably extract features from the measured part. Robust 2-D feature extraction requires high reliance on part presentation and contrast consistency.
When 2-D systems are used to measure complex assemblies like transmission snap rings or piston pin cir clips, many factors can affect part appearance and presentation, making the use of feature based measurements alone a less robust solution. Similarity in component color and finish as well as changes in metal reflectivity, or oil and debris on parts can negatively influence a 2-D system causing missed measurements and higher rates of unwarranted rejects (false positives).
Due to the rich set of data generated by 3-D systems, another set of virtual measurements can be added to the physical measurements to make the assembly verification solution more robust and less susceptible to the conditions that affect 2-D vision. Virtual measurements include feature centers, feature intersections and “fitted” surfaces that can be extracted from the full 3-D surface data that is generated. For example, on a typical transmission snap ring in a clutch housing, 3-D data will allow for the full inner and outer surface analysis, including inner and outer diameters, cross sectional sizes and roundness of the entire ring and housing surfaces.
A 3-D model of the as-built part can also be created, allowing for visual analysis of the full snap ring assembly. These virtual measurements can be combined with physical measurements to create an extremely robust solution with a high degree of redundancy built in.
Expanding the Market for Machine Vision ApplicationsThe focus of this article has been about applying 3-D machine vision to existing applications to make them more robust, simpler to use and lower cost to own. An additional benefit to the adoption of 3-D technologies and techniques for assembly verification is the expansion of the machine vision market for the in-line inspection of complex assemblies that have traditionally been handled by other methods like mechanical probes and sweepers or human inspection. Generating a rich data set that is less susceptible to issues created by part presentation changes creates more opportunities for machine vision to be used as a solution for the assembly verification of complex parts. Manufacturers will be able to reap even more benefits of machine vision in their facilities as 3-D further expands this growing market. Based on the apparent benefits, when evaluating solutions for assembly verification in today’s manufacturing facilities, 3-D is truly a better place to start. Q
For more on optical inspection, visit www.qualitymag.com :
Advanced Optical Technology Opens New Doors Advanced Optical Technology Opens New Doors
Find the Right Optical Option
Understanding Optical Inspection
Benefits3-D systems can offer intuitive answers based on 3-D models of the part built up from the part geometry itself.
3-D benefits are particularly valuable when vision is used to inspect complex assemblies, where the assembled part often results in several distinct shapes.
The part topography can be said to create “3-D relief” that can be collected in 3-D and used for multiple processing of the complex geometry.