The setup operation is a critical step in any cutting or forming operation. On one hand, the workpiece must be set up accurately to achieve the correct dimensions, but on the other hand, it needs to be efficient to maximize productivity. Setup errors are one of the main causes of producing out-of-tolerance parts. At the same time, the economic impact of the setup operation is significant because it involves both direct labor and machine time. This article describes a recently developed system that uses optical imaging and machine vision techniques to update the program data by locating known, visible features within captured images.
There are three main methods of setting up the workpiece to account for the uncertainty in positioning with respect to the data used for the cutting or forming operation. Using feeler gages between the workpiece and cutting or forming tool to sense the gap is the most economical, and perhaps the most widely used, technique. Once the gap is measured, the CNC program can be adjusted to account for the positioning uncertainty of the workpiece. Obviously, this is a time-consuming and crude technique that is usually limited to correcting the offset in only one direction.
The second most popular technique is to use fixtures to position the workpiece at a known and repeatable position. The main drawback of this method is the cost of designing and deploying special-purpose fixtures. For this reason, this method is usually limited to high-volume manufacturing.
A third method involves using mechanical touch probes to find the position of the workpiece. Conceptually, the use of mechanical probes sounds reasonable, but in practice, their use is limited for a variety of reasons. One is the absence of useable contact points or surfaces related to the section that needs to be cut. Another obstacle is having a machine that does not have the necessary hardware to interface with an off-the-shelf touch probe.
Machine Vision Techniques As A Setup Option
An alternative to the above-mentioned methods is to use machine vision techniques to optically image the workpiece. These techniques can also be employed to obtain the location of visible features such as holes, edges or corners to compute the exact location (and orientation) of the workpiece. In the past, using machine vision techniques to set up the workpiece was regarded as impractical. This was mainly because of the lack of suitable hardware and reliable, general-purpose pattern matching techniques to locate the features of interest. For example, up until a few years ago, solid-state cameras were either too expensive or too bulky to be packaged into a suitable scheme to deal with the harsh machining environment.
However, video cameras have shrunk in size as a result of advancements in digital imaging technology. A complete CCTV camera can now be fabricated using just one chip no larger than a fingernail. Computing power and computational techniques have advanced to such an extent that smart imaging sensors can be easily set up to perform complex detection, inspection or measurements.
The new system from OptoVue (Meridianville, Alabama) incorporates a CCD imager with a resolution of 640 by 480 pixels to view an area of 6.4 mm by 4.8 mm, thus resulting in an imaging resolution of 10 microns. The illumination source and optical element are both custom-designed and intenerated into a package that fits into a standard toolholder. This means that the user can simply obtain a standard toolholder and fit the imaging head into the toolholder as if it were just a cutting or forming tool.
The signal from the imaging device can be sent to any video input device, such as a frame grabber, VCR or display monitor. The video signal is connected to a frame grabber that is interfaced with a standard personal computer to acquire, process and display the images. This allows users to simply mount the workpiece on the machine's table at any position and orientation. The imaging system can then be used to find the exact position of the workpiece to rapidly reconfigure the CNC program.
Testing The System's Effectiveness
Two different experiments were carried out to assess the effectiveness of the imaging system in a shopfloor environment. With the first experiment, the main objective was to find the type of problems that could be addressed by acquiring and displaying the image of the workpiece. Upon consultation with expert machinists, it was determined that if the system could grab an image of the workpiece and superimpose a crosshair at the location that corresponds to the center of the tool (spindle), then a large number of problems encountered on the shop floor on a day-to-day basis could be addressed.
A calibration procedure was devised to find the geometrical relationship between the center of the tool (spindle) with the optical axis of the imaging system. This involved fabricating a small hole (in this case, 1 mm in diameter) at a suitable location within the work envelope using a tool that had little or no wobble. The depth of the hole is not critical. Hence, the selected tool can be mounted in the toolholder so that the wobble is minimized or, ideally, completely eliminated.
Upon fabricating this hole, the toolholder was removed (without moving the spindle or the table) and the imaging system, which had already been mounted into another toolholder, was mounted on the spindle. The image of the fabricated hole was then taken by the imaging system, and the offset between the center of the hole and optical axis of the system was measured. A graphical marker, such as crosshair, was then superimposed at the same location as the center of the fabricated hole. In other words, the location of the crosshair indicated the exact location of the spindle.
With the imaging system calibrated, the machinists were asked to evaluate the usefulness of the system. They came up with a number of cases in which this concept could be beneficial. For example, one machinist used the system as an alignment tool for rapidly setting up the workpiece. Another used the system to turn his vertical milling machine into a simple coordinate measuring machine (CMM) by aligning known visible features with the crosshair and recording the table position (which is the same as the spindle position) to compute the required dimensions. Another machinist had to drill small holes at a certain angle with respect to another feature for fabricating an advanced rocket engine. This machinist used the imaging system to assist him with the cutting operation.
With the second experiment, the main objective was to study the requirements for achieving fully automatic setup of the machine irrespective of the workpiece position and orientation. Two key issues were investigated. The first was to identify what physical features are needed to perform automatic setup. The second was to decide if the CNC program should be reconfigured to account for the positioning uncertainty of the workpiece, or if the workpiece should be translated and rotated to align its axes with those of the machine.
The first issue was relatively easy to address. The exact position and orientation of the workpiece could be computed from the location of two (or more) known points, a point and a line, or two lines. The points can be corners, center of holes/arcs and so on. Similarly, the lines can be the edges of the workpiece. Once the amount of translation and rotation of the workpiece are known, a simple linear transformation can be applied to either correct the workpiece location or reconfigure the CNC program.
To address the second issue, a sample of typical CAD drawings was studied. The study concluded that the choice between correcting the part position and orientation versus reconfiguring the CNC program was either application-dependant, or it did not matter at all. The critical factor was whether the cutting or forming operations involved coordinated movements of several axes, such as following the contour of an arch. If they did, then there was no difference between translating/rotating the workpiece versus reconfiguring the CNC program through a linear transformation to follow a different path.
There was no significant advantage in either method when the operation involved moving to adjust a point and then cutting or forming, such as when drilling or boring holes. On the other hand, when coordinated movement was not required and the movement of the tool was in one direction only, moving the work had an advantage because one can rely on the mechanical rigidity and accuracy of the axis to guide the tool along a straight path as opposed to performing coordinated multi-axis moves to achieve a straight line path.
In addition to the benefits resulting from the rapid setup of the workpiece, the optical imaging system can also improve productivity and quality by enhancing machine use. For shops that deploy a limited number of machines, there is usually little or no choice as to which machine is used for which job or part of a job. However, with larger organizations that have a diverse selection of machines, one can select a particular machine for a set of operations to achieve the optimum machine use.
For example, a complex job may require large but not accurate cuts as well as small and accurate ones. In this case, the operator can distribute the load across several machines so that the large cuts are performed by a heavy-duty machine that is fast, but not necessarily accurate, while using a lighter and more accurate machine for the more delicate and high-accuracy cuts.
Of course, this concept is not new, but its use has been limited because the setup process, a time-consuming operation, must be duplicated on every machine, and the buildup of tolerances from multiple setup operations may result in unacceptable overall quality. The concept has been used within the electronics industry for many years; each machine along the manufacturing line performs a section of the complete assembly operation. To achieve the same methodology in the machining industry, material registration by virtue of fiducials, which are commonly used in the electronics industry, needs to be carried over to the machining industry. In the electronics industry, a set of fiducials (at least two) is printed on each printed circuit board (PCB). At each machine along the manufacturing line, the locations of the fiducials are obtained through machine vision techniques, and the exact location of the PCB is computed.
A similar process can be applied to the machining industry. Distinct visual features, such as round shallow holes, are cut into the workpiece at the first station that performs the initial cutting or forming. Subsequent machines can then be equipped with the optical imaging system to find the location of the fiducials and compute the exact location of the workpiece. This information can then be used to either correct the position of the part or update the CNC program, whichever is applicable.
In addition to easy and efficient setup, the optical system can help achieve in-process quality or process control. The location of visible features can be obtained by moving the imaging head over the required location and capturing an image of the workpiece. The exact location of visible features can be computed from the camera location (in this case, the spindle or the table location) and the image location of the visible features. Because the images are taken at high magnification, the measurement resolution is significantly fine to facilitate accurate measurements.
Traditionally, there has been a resistance to using a cutting or forming machine as a CMM. Although off-line dimensional gaging will always have a place within the machining industry, this traditional resistance is gradually giving way to modern manufacturing practices. The availability of the optical imaging system described here, as well as advancements in machine vision software and computing technology, offer justifications for the machining industry to incorporate some form of optical in-process quality control to improve yield and productivity.