(Note that Nissan denied that a goal of high as the expert system did. This cutting All items designed in NADAMS are in one database accessible to the designers, including those who design production equipment. Casting and molding dies, NC machine operations, robot programming, operating models of parts and products, and simulations are example applications supported. There is an expert system to help devise cutting process plans, typical CAE for vibration, stress and thermal analyses, mold flow simulations to aid die design, and some fault tolerance analysis software that was not explained further. For example, the mold flow program (IMAP, developed by Toyota Central R&D Laboratories, Inc.) helped Nippondenso reduce the weight of its air conditioner case and avoid having a hole develop during molding. The number of actual prototypes needed was reduced by 66%. The metal cutting expert system is based on Metcut's data plus 650 rules provided by Nippondenso's process engineers. The rules comprise knowledge about how to process certain geometries plus formulas for calculating feedrates and tool wear, for example. The software chooses tool material and size, cut depth, feedrate, cutting speed, and cutter rotation rate. In a side-by-side test, process engineers provided process plans for a precision surface that varied by 4 to 1 in recommended cutting speed. Only one engineer recommended a cutting speed as The facility I toured was a training center. It contains a wide variety of workstations but mostly IBM 5080s. I saw two demonstrations: robot offline programming and supercomputer output showing FEM studies. Robot offline programming is supported by a wireframe 3D modeler that permits a user to build up a model of a workstation from basic shapes. A primary function of the program is to predict and improve the cycle time of the robot workstation. The computer already has models of Nippondenso's various robots (which it makes in-house). I could not find out how the coordinate data were put in so that workpieces, fixtures, and teach points for the robot could be described. Collision avoidance is done by trial and error, using the modeler's intersection capability. Straight line paths are computed automatically as a first try and the user modifies them to avoid obstacles or improve cycle time. Several FEM examples were available. These include fluid flow in plastic injection molding, turbulent mixing and heat transfer inside the air conditioner between cold and hot air, stress-strain, and flow inside a fuel manifold. NADAMS supports pre- and postprocessing, and a commercial FEM package does the calculations on the mainframe. Developments in Two interesting activities of the by Mr. Harada and Mr. Sugito: design for assembly (DFA) and assembly technology. The Assembly R&D Group has only five members and was begun in 1985. Its jobs include interacting with the research community at home and overseas, developing ways to simplify products using their own DFA methodology, and developing ways to assemble difficult products that can't be simplified. Assembly technology is divided into two parallel efforts: dexterous/intricate assembly and large variation assembly of simple items (Figure 4). Engineering innovation is used on the first kind while economic approaches are used on the second because they are already technically easy but too costly to automate. For large variation products, an economic analysis showed that cost of preparing and feeding parts grows much faster than other costs as the number of variations grows. Efforts are going into various "low cost" feeding and preparation methods, including an attempt at low cost bin picking. Bin picking is being used in only one factory application, however. Other applications are under development. Reconfigurable grippers and pallets are also under consideration, along with such approaches as molding groups of parts onto one backbone and cutting them off at the moment of assembly. For technologically challenging assembly tasks, such as fitting unwieldy, flexible, and warped items together, Nippondenso long ago concluded that "intelligent, dexterous, and adaptable" robots were too expensive or unavailable. Instead, they decided to “utilize the characteristics of the product" as well as to redesign the product so that assembly could be accomplished. This is another example of the "smart product" approach. A fine example shown was fitting top and bottom halves of molded plastic air conditioner housings together (Figure 5). These fit by tongue and groove around a large perimeter ("island"). Since the cases warp, the halves cannot just be pushed together. Fixturing could be used to force the halves into the correct shape but that would require costly fixtures and/or making the parts too flimsy. The problem gets worse when the joint has a gap ("discontinuity”) or two rather than covering the entire perimeter. The worst situation occurs when there are "intermediate parts" such as pivoted damper doors where one end of the hinge pin fits in a hole in one case half and the other end fits in a hole in the other half. Such doors are placed upright in the lower half but flop over to one side and the hinge pin will not line up with the with the upper half's hole. People currently assemble these parts. They push and bang the case halves together, reaching inside to line up the damper hinge pins and the case holes. It is an obvious bottleneck on the production line and inherently difficult to automate. The robot solutions have been demonstrated in the laboratory but not applied in the factory. They are elegant and involve a mix of robot angular maneuvering of the top part, redesign of tongue and groove shapes, and redesign of damper doors. This is the approach I called "smart product" above. To fit a tongue-groove that covers the entire perimeter, the robot tilts the top half and mates the parts on one side. It then pivots the top half down gently by hinging at the initial contact point, and the tongue rolls into the groove. When the joint has a gap, the above method is used, starting at a pivot point opposite the gap and rolling around so that the parts are mated at one end of the gap. A vision system is then used to find the top in relation to the bottom at the other end of the gap, and the robot pushes and slightly deforms the top half until the parts are aligned. Then the pivot-roll method is used to mate the parts while not disturbing the mate achieved at the first end of the gap. When there are several gaps, the one in the most flexible region of the case is mated first, then the next most flexible, and so on. When there is one damper door, the robot pushes it upright with the top half of the case and catches the door hinge in the hole in the case. Then it repeats the tongue-groove method. The hinge pins on the damper are made extra long so that they do not fall out during the pivoting operation. When there are several doors, this process is repeated for each door, and the hinge pin of each door is designed to be longer than that of the next door so that the sequence of door mates can be controlled. Whether this scheme can be applied reliably and at high enough speed in the factory is unclear at this time, but given Nippondenso's past record, it will be. It is a pretty sophisticated approach and represents "design for assembly" as high art. fixed portions and variable portions, suppressing minor variations and using more common parts instead, and using the FMS-1 technique. This topic is a subject of ongoing research and Harada is gathering more examples from around the company. Twice a year they hold a DFA seminar to trade stories, hear advice from both product engineers and process engineers, and teach the method. Harada's goal is to create a DFA program based on a solid modeler that will help product designers evaluate their own designs. Other companies I have asked about such an approach (a subject of my own research) say that they do not believe product designers will ever have the time or knowledge to do such evaluations themselves. Harada will move to Nippondenso Technical Center U.S.A., Inc. near Detroit and will survey research opportunities from there. Kimura noted later that both Boothroyd/Dewhurst (B/D) and Draper Laboratory research on design for assembly and simplification of products has had a strong influence in Japan. The B/ D method is very popular although its limitations are recognized. Nippondenso has also developed its own DFA evaluation method. TOYOTA Nippondenso's method is broader and more sophisticated than typical DFA 31 July 1991 methodologies, which most people agree focus too much on small parts. It con- Background tains 65 points of evaluation, such as how parts must be prepared for feeding, how many variations there are in parts and product, whether a part's feeding method supports variety, how difficult the assembly technique must be, and how many parts there are (the most important item). Production engineers perform the evaluations and give advice to the product designers. An interesting redesign activity is called variation reduction. Its aim is to reduce the effect of multiple models on the assembly processes. Methods used include modularizing the product into My host was Mr. Y. Kuranaga, Head of Development Div 1 of Information Systems Div 1 in the Information Systems Division. The entire operation has 641 people plus at least 330 outside software developers. Its mission is to develop and disseminate software for engineering, business, and factory operations, plus to provide software training. Toyota has agreements with several companies such as Nihon Unisys to support and sell its CAD software to Toyota's vendors. |