Patent Application
20070186417
This invention relates generally to a method for electronically identifying coded manufactured parts, and more specifically, to a method for identifying cast alloy wheels on-the-fly during the manufacturing process. The invention further provides a novel, distinct solution for tracking vehicle wheels and other parts through various stages of manufacture.
It is important that certain manufactured parts be traceable back to and through manufacturing. Thus, for example, if a safety critical defect is detected in a particular part, then parts with the same history can be identified and tracked down to minimize damage and liability. As such, it is essential that many parts be permanently marked in such a manner that manufacturing records for the part can be retrieved for analysis at an unknown future date. For example, aluminum wheels are a primary structural component of an automobile that requires full traceability for safety reasons.
Reliable marking and automated scanning into a manufacturing database enables wheel serialization. A particular benefit is that scanning the serial number provides the exact wheel model and prior history, which is often necessary to adjust the upcoming process equipment during manufacture. Currently, this is either done with operating staff or machine vision systems that look at the face of the wheel. For some processes, it is also important that the angular orientation of the wheel is also identified.
The manufacturing of cast alloy wheels is generally an ordered process of sequential events. Some of these events are specific to an exact wheel model, while others are not. For example, machining is geometry specific. If the wrong casting is loaded in an automated machining cell, a dangerous and expensive crash occurs. Heat-treating, on the other hand, is non-specific to the wheel model. However, as heat-treating is approximately a single shift long process, it is still useful to know what is in the furnace for planning the subsequent operations. For reasons as diverse as these, it is advantageous to identify wheels during manufacture by their model number.
The most common method to identify wheels is by a human operator. But in higher volume automated operations, this is both expensive and less than 100% reliable. Consequently, sensor-based wheel model recognition systems are desired. Various sensor technologies are used, the most prevalent being machine vision. Here, a snapshot of the wheel face is taken and compared against stored values. While this is generally straightforward for a human, it is a difficult task for machine vision, primarily because the snapshot is only a 2-D image. Often such systems are only useful when other inputs are used in parallel, or series snapshots are required to eliminate the probability of misidentification.
The need for automated rapid identification of manufactured parts has led to information encoded as 1-D bar codes for machine-reading. These all-pervasive linear barcodes are typically high contrast marks, most often black bars on a white background to facilitate reliable and rapid scanning and decoding. When low contrast barcodes are used they are generally unreliable with conventional scanners. A solution used to overcome the low contrast issue is to use a particular type of 2-D barcode, where the bars are either below or above the general surface. Then, by using more sophisticated scanners, for example, those based on laser distant measurement units, these linear bar codes can be reliably read. This type of 2-D barcode is generally referred to as linear “bumpy barcode.”
For hostile and abrasive environments, the use of bumpy barcode and other 2-D DataMatrix like area barcoding formed by DPM processes is common. While such laser formed area marks are more or less 2-D, peened area marks are actually 3-D marks. Either way, low contrast is often encountered, and such marks can only be scanned satisfactorily when special contrast enhancing lighting and narrow field of views are practiced. Eliminating surface variations in the area carrying the mark and controlling its orientation relative to the scanner are often required to improve automated read rates.
Scanning Background
Direct part marking (DPM) with peened codes presents a visual contrast problem for scanners. Suitable lighting resolves this constraint for many parts; however, round parts such as wheels pose particular difficulties, especially when on-the-fly scanning is required.
In general, it is advisable to have the entire code in the scanner field of view (FOV) for robust decoding. This is particularly true for position-based area and bar codes. Some code variants with a timing mechanism to desensitize positioning can sometimes deviate from this FOV requirement, especially if the coding alignment and orientation are constrained relative to the lighting and scanner. Also, if the scanner resolution and scan rate is high enough relative to any motion irregularities, the code does not necessarily need be in a single FOV.
Peened area codes are in the broadest sense three-dimensional coding that can be detected, for example, by tactile surface scanning. They also can be detected by non-contact surface scanning, such as by optical triangulation or interferometery methods. All such advanced optical systems require specialized illumination of the code. More conventional vision based systems also require controlled lighting to enhance the peen code contrast.
Laser code scanning is used for high contrast printed bar codes, but it is not practiced industrially for peen coding. To be suited for peen marked parts, laser surface mapping is required. Then, by subtracting the background surface, the code can be “seen” for decoding. While other techniques can be used, single point laser distant measurement based on triangulation is the basis of surface mapping where there is a large FOV and DOF requirement, as needed for wheel code scanning on-the-fly. While the single point can be upgraded to a continuous line without raster movement, the complete area code can never be in the FOV. As such, the peen coding must be sufficiently pronounced to prevent relative movement sensor “noise” from losing or adding individual peen marks and locations. Fortunately, this does not appear to be a problem for marking wheel castings for conveying past a fixed-mount scanner, as they have overly large machining stock on the desired rim locations to facilitate castability.
The broad concept of the present invention is to mark a manufactured part, such as a wheel casting, with a machine-readable area relief pattern. In one application, the area relief pattern is a 3-D peened code—although other coded surface profiles are contemplated. The wheel can be scanned on-the-fly to generate a 3-D surface map in a region of interest containing the coded pattern. By extracting the coded pattern from the measured region of interest, a suitable decoder can then extract the part information contained in the code.
Therefore, it is an object of the invention to provide a method for electronically identifying a coded part.
It is another object of the invention to provide a method for electronically identifying a coded part on-the-fly during manufacturing.
It is another object of the invention to provide a method for electronically identifying a coded part on-the-fly as speeds in excess of 1 foot per second (fps).
It is another object of the invention to provide a method for electronically identifying a coded part which utilizes precise and accurate means for marking and reading a coded 3-D area relief pattern.
These and other objects of the present invention are achieved in the preferred embodiments disclosed below by providing a method for electronically identifying a coded part. The method includes the steps of locating a machine-readable area relief pattern formed with a surface of the part. The relief pattern comprising separate and distinct code elements extending along both x and y axes. Each code element has a profile dimension extending along a z-axis relative to a native surface of the part. A region of interest containing the area relief pattern is then measured along the x, y, and z axes. The area relief pattern is then extracted from the measured region of interest. The area relief pattern is then decoded to extract part information encoded in the relief pattern.
According to another preferred embodiment of the invention, the step of measuring the region of interest comprises employing a laser line scanner adapted for projecting a laser line onto the surface of the part containing the area relief pattern.
According to another preferred embodiment of the invention, the step of measuring the region of interest further comprises moving the coded part relative to the laser line scanner.
According to another preferred embodiment of the invention, the method includes measuring the region of interest on-the-fly as the coded part is moved past the laser line scanner.
According to another preferred embodiment of the invention, the method includes moving the coded part past the laser line scanner at a minimum rate of 1 fps.
According to another preferred embodiment of the invention, the method includes arranging multiple laser line scanners at predetermined locations relative to the moving coded part.
Preferably, the area relief pattern comprises a peened area code.
According to one embodiment, the coded part comprises a cast alloy wheel.
Preferably, the area relief pattern is formed with a rim barrel of the wheel.
Some of the objects of the invention have been set forth above. Other objects and advantages of the invention will appear as the description proceeds when taken in conjunction with the following drawings, in which:
Referring now specifically to the drawings,
The wheel 10 is generally processed after casting in a face-up position with the inboard flange 15 resting directly on a powered roller conveyor “C”, as best shown in
Standard line of sight tracking systems require the wheel identification mark to be presented within the scanner's field of view. While several non-contact distance reading technologies are suitable, to get high resolution the distance variation of the mark to the scanner (depth of field, or DOF) must be kept within a relatively tight range—usually under 50 mm. The field of view (FOV) of such high-resolution scanners is also relatively limited—in the sub 100 mm range. The DOF and FOV of optical camera vision scanners are significantly less—especially the DOF. The wheel identification mark, which can be any size but is typically in the 10 mm range, is preferably read when perpendicular to and in the same plane as the scanner.
As indicated above, the concept of the present method is to locate and read the wheel identification mark “M” on-the-fly during processing without slowing or stopping downstream forward movement of the vehicle wheel 10.
Referring to
Preferably, the code reader or scanner “S” discussed above comprises a non-contact laser line scanner such as that offered by Micro-Epsilon of Raleigh, N.C. under the name LLT 2800. As demonstrated in
The resulting output is a data stream that creates a point cloud. This point cloud contains not only the planar data of the measured surface, but also the depth; it also includes various deviation errors. In order to convert this point cloud to a planar surface, a certain data handling approach is required to correct the errors. First, the line data is sorted, both the line point data set and the line to line sets. Then, the cleaned 3-D point cloud is converted to a quadrilateral meshed surface volume, so that precision surface modeling with NURBS, for example, is performed. The next step is to correct for shape deviations, the main influence being the curvature of the cast outer wheel rim surface radii. Once this is finished, a flat equivalent surface map results that contains in one portion the peen-coded region. Next a region of interest (ROI) algorithm is applied so that only the coded surface region is analyzed to limit the final decoding task.
The decoding task involves determining the peen marks and their rough relative positions, and then using defined data matrix layout rules and geometry to reconstruct the peen code on a perfect grid pattern. At this point data matrix algorithms are applied to determine if the code is a decodable matrix, and then to extract the coded information contained in the matrix. This information is then available for further use, such as an alphanumeric display or insertion into data bases.
A more complete and detailed discussion regarding laser point cloud analysis and mathematical manipulation is provided in the article Evaluation and Correction of Laser-Scanned Point Clouds by Christian Teutsch, Tobias Isenberg, Erik Trostmann, Michael Weber, Dirk Berndt, and Thomas Strothotte, and published in Proceeding of Videometrics VIII (Electronic Imaging 2005, Jan. 16-20, 2005, San Jose, Calif., USA), volume 5665 of SPIE Proceedings Series, pages 172-183, Bellingham, Wash., 2005. SPIE/IS&T. The complete disclosure of this article is incorporated herein by this reference.
A method for electronically identifying a coded part is described above. Various details of the invention may be changed without departing from its scope. Furthermore, the foregoing description of the preferred embodiment of the invention and best mode for practicing the invention are provided for the purpose of illustration only and not for the purpose of limitation—the invention being defined by the claims.
| Number | Date | Country | Kind |
|---|---|---|---|
| PCT/US04/18082 | Jun 2004 | US | national |
| PCT/US04/23133 | Jul 2004 | US | national |
| Filing Document | Filing Date | Country | Kind | 371c Date |
|---|---|---|---|---|
| PCT/US05/06898 | 3/3/2005 | WO | 9/1/2006 |
| Number | Date | Country | |
|---|---|---|---|
| 60549822 | Mar 2004 | US |
