Manufacturing insights

Manufacturing insights

Journal of Manufacturing Systems Volume 5i No. 3 book review S.J. Williams et al. of the Cambridge University present under the title "Faster and Mo...

264KB Sizes 2 Downloads 91 Views

Journal of Manufacturing Systems Volume 5i No. 3

book review

S.J. Williams et al. of the Cambridge University present under the title "Faster and More Accurate R o b o t s - The Impact of Advanced Control Design" their work in control "topology" and in computer-aided design of robot servo-systems. They claim that the"star" scheme of signal transmission from the master processor to the independent axis servos introduces a communications bottleneck which slows down the robot. A better scheme--ring based with parallel processors is described. New control algorithms and trajectory "feed forward" control improve the dynamic accuracy of the robot. The CAD system they use for servo design is very extensive, the result of 10 years of development work.

allows the arm to enter a tube of 212 mm diameter. The payload is up to 35 kg, work space--a hemisphere of 1200 mm radius, tool tip repeatability is 0.5 mm. The manipulator is being commercially produced. (The review of the proceedings will be continued in the next issue.)

--Professor Moshe M. Barash, Associate Editor

Machine Vision Issue No. 4 of Manufacturing Insights, Videotape Series

J.T. Meij and C.J. Franken of the University of Stellenbosch in South Africa present an"Economic Analysis Model for Robot Applications" which can be run on a personal computer by managers and analysts with little knowledge of computers. Many variables, such as human resources, inflation, multirobot workstations, multiple shifts, etc., can be "factored in", and the effect on scrap reduction, reduction in inventory, savings in material, in labor, etc., etc., is obtained, all in spread sheet format. Sixty-five variables and parameters are listed, and sample analyses are presented.

Society of Manufacturing Engineers, © 1985 45 minutes $200.00 The fourth in the Manufacturing Insights series of viedotapes intended for manufacturing management executives provides a condensed overview of machine vision, especially as applied to manufacturing industry. Machine vision is an intriguing term to the uninitiated; obviously it cannot be the same as television because there is never the mention of a "machine" in connection with the latter. Of course, come to think of it, most of use have seen examples of"machine vision" in such movies as "Star Wars" of recent days and "Metropolis" of half a century ago. These were the seeing robots--whether R2D2 and C3PO in the first or the rebel woman worker in the latter. But this was in movies, not "for real".

Returning to the engineering aspects, R.K. Stobart of Cambridge Consultants, U.K. talks about "Using Solid Modellers in Robot Programming". In short, examples of modeling robot movements with the BUILD package are presented. An example of robot application to processing plastics is given by W.G. Bryce and J.P. Watkins of NEL in a paper entitled "Robots and Fibre-reinforced Plastics: Preparing the Charge for the Moulding Press". The product was a small car wheel, made of fiber reinforced plastic. The paper describes in some detail the process in all its stages and the role the r o b o t - - P U M A 560--has in the process. The manufacture of the wheel is quite complex and at the time of writing, lay-up of the charge was not yet automated.

Rudiments of actual machine vision are, however, much older than most of us realize--these were the early "electric eye" signaling systems that counted visitors in a museum or protected a store from burglars. Their level of intelligence was, of course, rather low, about that of an ameba. M o d e r n machine vision dates back to the marriage of the TV camera and the computer, some 30 years ago. Even the old TV camera of those days was not too bad for many industrial purposes, but the computers were hopelessly slow and limited. It took over 20 years of progress in computer capability and of research in mathematical methods of image processing to develop the first practical industrial vision systems. Since then the rate of advance accelerated. Microcomputers of major power (number of transistors) and higher speed, cameras with more picture elements (pixels), and more efficient mathematical image analysis methods, have brought machine vision to its present level. A cross section of modern machine vision applications is presented in the videotape.

The paper that followsw"Latest Developments, by Taylor Hitec Ltd., in Power Manipulators and Deployed Robots for the Nuclear Industry and Elsewhere" by D.B. Lowe of Taylor Hitec Ltd., U.K., describes a mechanical "elephant's trunk" called "The Advanced Manipulator". The purpose of this device is to perform precision work, such as welding, at a considerable distance from the controller. The arm of the manipulator is articulated, and had undergone two redesigns. The actuator motors and transmissions are housed inside the tubular limbs which

214

Journal of Manufacturing Systems Volume 5/No. 3

book review

After an introduction into the basic concepts of machine vision, four experts from leading machine vision companies are interviewed, each dwelling on some important aspect of this technology.

Managers viewing the tape should become convinced that machine vision is a real, viable technology and that engineers wishing to employ it should be given a chance to do so.

The major part of the video presentation are five case studies, four from manufacturing and the fifth from... agriculture...

The plant produces 30 million crowns a day, and another in Oklahoma has similar output. Each plant has nine lines, every one equipped with the vision system. The sampling of 400 million crowns in the first quarter of 1985, with the new system operating, revealed 11 defects. The actual number of defects per machine per shift was reduced from 750 to 7-11. The reliable inspection made it possible to speed up production by 24%, and the savings because of reduced scrap amounts to $75,000 annually. Soon another camera will be installed to check the color printing on the crown, and a new system will be developed for inspecting aluminum closures. The above system is an example of the high speed attainable, in appropriate applications, with machine vision as compared to human.

The first case study is the General Motors Assembly Center in Orion, Michigan. Here bodies of "high class" Oldsmobile and Cadillac two- and four-door cars are assembled and welded. Until 1984 the assemblies prior to welding were manually inspected which was difficult for the workers and often required correction. A computerized, flexible, optical on-line inspection system for the various body styles was installed, in which 160 test points on the body are checked to 0.1 mm repeatability by noncontact methods in 25 seconds. The precision of the system, bearing in mind the size of the workpiece, is most remarkable. The numerous sensors, which use laser and unstructured light, are of four general types--one-axis, two-axes, and three-axes with object discrimination from background (e.g., recognizing a stud on a surface and determining its location). The system generates on-line information which is used to correct the manufacturing process if deviations approach limits. The assembly quality has improved, and with it, customer satisfaction. Although individual cameras are not very complex, their large number combined with well thought out total system design have produced a most impressive result. Additional "machine vision" inspection systems for car body assembly are to be installed in the plant.

Inspection of printed circuit boards for critical equipment in aircraft, space and missile done at Kearfott Division of the Singer Company in San Marcos, California is shown next. A sophisticated machine vision system is employed which has pattern recognition capability. With it, both circuit board artwork as well as actual plated and drilled boards are inspected. Various defects, such as wrong hole diameter, faulty conductors, extraneous copper, etc., are automatically detected. The automatic inspection is eight times faster than visual (human), and the immediate feedback permits quick process correction. Shown next is another automotive application--this time at the General Motors BOC Group plant in Lansing, Michigan. The addition of four-door models to certain two-door Oldsmobile, Pontiac and Buick cars made it necessary to build new door welding facilities. Whereas fixed automation is used in the older welding systems for the two-door models, two flexible systems were designed for welding the new doors. Both systems integrate welding with laser based machine vision. Eight welding robots work in the front door system, six in the rear door system.

The second case is the inspection of a humble product--the metal cap or "crown" that seals beverage bottles. The manufacturer, Zapata Industries in Frackville, Pennsylvania, wanted to achieve cap inspection that is faster; more reliable, and preferably cheaper than manual inspection. The industry standard permits one defect per 10,000, which was increasingly difficult to achieve by human inspection as manufacturing line speed gradually increased from 1600 to 8400 crowns per minute. A noncontact vision system with a solid state camera was developed and installed. It operates with infrared light and can check for 24 possible defects (quite a lot for such a simple item!).

The first system is described in some detail and employs 18 cameras which "track" the parts, inspect the gap between the parts to be joined by welds, namely the door panel and the "header" (the window frame), and guide the welding robots. The entire welding operation is performed on a moving line to save time. Operating in the "just-in-time" mode, the line has sufficient stock for just one hour of work. The welding of a pair of doors takes only 30 seconds. The welds are automatically inspected

As before, the SME has produced an excellent tape. Manufacturing engineers should view it carefully, several times over. It shows the great diversity of problems that can be solved with the aid of computer vision, and offers an inspiration to an engineer having to automate an "awkward" process.

continued

215

Journal of Manufacturing Systems Volume 5/No. 3

book review

and defects immediately reported so that the process can be corrected.

provides raw material for some 1500 byproducts. Manual inspection and sorting required 55 persons. The new system, which together sort 100 lemons per second, require only 6 to 9 persons, and the sorting quality is better.

The last "case study" is reported from the lemons sorting facility of the Seaboard Lemon Association in Oxnard, California. Here, lemons are inspected and graded. In the grading room, ten sorting lines carry washed lemons, 2.5 million per 8-hour shift, through sophisticated inspection systems. First, the lemons are x-rayed, and those with air pockets are rejected. Then, each lemon is viewed simultaneously by eight cameras, four of which check the color, the other four check for surface blemishes. Lemons are sorted into four colors: yellow, silver, green, deep green. Blemishes as small as one mm diameter are detected, the total blemished area of each lemon is automatically computed and the fruit accordingly sorted automatically into the "Sunkist" grade, unnamed fresh lemons, and "culled", which are processed into lemon juice. The rind of these lemons

The video presentation is concluded with expression of opinion about the future of machine vision by several experts. There is no question that machine vision is advancing fast technically and finding ever more new application areas. As an industry, machine vision is expected to grow at the rate of 35% annually, and reach sales of one and one half billion dollars by 1994. The trend will be toward complex systems in which machine vision is integrated with other advanced, computer controlled technologies.

--Professor Moshe M. Barash, Associate Editor

BOOKS IN REVIEW

Note: Books listed in this section are under consideration for subsequent review.

Electronic Weighing in Industrial Processes K. Elis Norden Granada Publishing Ltd., London, © 1984 ix + 300 pp. $55.00.

Manufacturing Automation Management Robert W. Bolz Chapman and Hall, © 1985 xi + 252 pp. $27.50.

Materials Processing with Lasers and Electron Beams

The Effectiveness of Flexible Robotized Manufacturing Systems (Effektivnost perenalazhivayemykh robotizirovannykh proizvodstv), V.A. Kozlovskii, et al. (in Russian) Mashinostroyenie, Leningrad, © 1985 224 pp. $3.00 (approximately)

(Lazernaya i elektronno-luchevaya obrabotka metallov) N.N. Rykalin, et al. (in Russian) Mashinostroyenie, Moscow, © 1985 495 pp. $6.95 (approximately)

216