PSU Magazine Winter 1988

to perfect a system which can automatical– ly analyze these pattern of points and decide if the part is in tolerance. On the simplest level this function would require a human inspector to physically fit the manufactured part into a pre-made gage to see if it is in tolerance. Right now in many plants, one in every 10 employees is an inspector. Labor costs can run high and effectiveness low because of the tedious nature of the task , and with the advent of new, hard-to– machine materials and precise tolerances, even the best human operators have reached the limit of their ability. If he is successful , Etesami's research could some day lead to more accurate and faster inspections without interruption to the manufacturing process. And the highly-accurate measurement data could be retained and used - possibly instan– taneously - to avoid product flaws in the first place. Computer manipulation of this data is Etesami's primary research focus, but his findings will not have much impact without a better measuring method than a microscope and positioning table. That is why he is working with a vision system. The system in Etesami's lab consists of a mounted video camera with the ability to freeze an image into a grid. This grid is fed into the computer where it is digitized into an "ant army of numbers ," says Etesami. Some day these numbers will be quickly analyzed using pro– grammed points of reference. Unfortunately, the resolution of this vi– sion system, and all such systems now in use, is not accurate enough for Etesami's purposes. Edges are fuzzy, so accurate coordinate points cannot be determined . In a parallel research project Etesami hopes to increase the accuracy of these systems by measuring from a "grey scale" showing gradual change of intensity around the edges of an object. If the in– dustry begins concentrating on more precise measuring, Etesami expects the equipment will improve. Manufacturers use vision systems now for inspection purposes, but it is usually to identify a part, see if it has three holes or four, if it is the proper length and den– sity, and if there are any cracks. "Nobody has attempted to get an exact point," says Etesami . PSU 10 Manufacturing in the precomputer era. In the automobile industry vision systems are used in many areas from finding missing holes in engine mounts, to recognizing the image of a car body and telling robots down the line what paint program to use. K nowing today's sophisticated side of manufacturing with its reliance on computers, vision systems and even robots, it is hard to believe that only 150 years ago mass production was first implemented . In 1855 Samuel Colt and Elisha Root had 1,400 machines producing precision weapons with interchangeable parts ·at their armory in Hartford , Conn . The two were building upon the manufacturing principles of Eli Whitney who promised in 1978 to produce 12,000 muskets with inter– changeable parts for the U.S. Government. Whitney took eight years to deliver, and historian debate today whether the parts were really interchangeable. Machine refinements made through the latter half of the 19th century led to the development of the first automatically controlled machine, C. M. Spencer's automatic lathe. Using what he termed a "brain wheel ," Spencer fostered the birth of automation. By the early 20th century some in– dustries had become so adept at turning out masses of parts that they encountered a bottleneck at the point of assembly. But the answer was found on the eve of World War I when Ford Motor Company developed the assembly line. The promise of mass production that began in the ar– mories at Springfield and Harper's Ferry was now in full swing. But manufacturing had yet to realize true automation in which computers are the foremen. In the late 1940s John T. Parson invented a way of guiding a mill– ing cutter to generate a smooth curve us– ing numerical control (NC). Coordinate points were coded onto punched cards that directed the machine to move in small incremental steps along the desired curve. Under US Air Force sponsorship MIT further developed this concept producing the first NC prototype in 1952, and in– dependent machine tool builders went on to develop different NC machines to meet their own particular requirements. When computers became available in the late 1950s, MIT was again sponsored to design a computer part programming language that could be used to describe geometric tool movements for NC machines. Today, CAM systems can automatically generate NC programs and simulate tool paths quickly on graphics display for verification. In addition , most systems can determine the sequence of fabrication steps, direct the flow of work and materials through the factory, and in some cases provide robotic manipulator arms to handle workpieces and tools. With such capabilities it makes obvious sense that inspection should also be a part of the CAM system. But the question in- · variably pops up: since these systems seem so perfect, why are flaws ever produced? "These are mechanical devices that are producing things," answers Etesami . "The accuracy of those systems is not infinite. There are positioning errors, tools wear (Continued on page 22)