Flexible system makes small lots of parts feasible
Automated system cuts set-up time and manufacturing costs for repeat orders of precision-machined optical-system parts.
Rosie Abriam
Precision-machined parts account for a high
percentage of the manufacturing costs of most electro-optical systems. These parts typically sell in volumes of a few hundred per year, and the small volumes translate into high manufacturing costs because machine shops have to amortize their set-up costs over only a few parts.
New Focus and its R&D subsidiary, Focused Research, have developed a rapid-response, flexible manufacturing system to produce small quantities of machined parts while eliminating recurring set-up time. With support from DARPA, we have devised a way to avoid setup costs for second and subsequent runs of a given part. Using our automated system, we can manufacture machined parts such as optical mounts, housings, and other components of optical systems in quantities of one as cheaply as typical job shops can manufacture those parts in batches of 25 to 75. The cost advantage for the second and each subsequent run of a typical prismatic part with six angular planes of work can be substantial for lots below 25 parts.
The conventional way
To fill an order for a six-sided prismatic part in the conventional way, an experienced machinist has to program the milling machine and assemble jigs and fixtures to hold the part, load the required cutting tools into a tool carousel, and position the fixtures in the machine. If the part requires machining on all six sides, the machinist will have to load it into the milling machine six times. Once the setup is established, however, the incremental cost of making the new part is minimal because the process itself is simple--the operator loads the part into the fixtures, and the machine cuts the part.
To fill subsequent orders for the same part, the machinist can retrieve the CAM (computer-aided manufacturing) program and the fixtures, but he or she must repeat the time-consuming set-up process--find the required tools, install them in the machine`s tool carousel in the same pockets where they were the first time, lay out the job manually (that is, calculate the work offset--the x-y coordinates on the raw aluminum plate--for each part), and enter that information manually into the keyboard at the console of the milling machine. In principle, the machinist could program the machine to make more than one type of part during the same run, but such programming would be prohibitively time-consuming.
To avoid recurring set-up costs, most machine-shop customers order machined parts in so-called economic order quantities--typical lot sizes of 25 to 200 pieces or more--stocking up on extra parts before they are needed and possibly having to scrap them if engineering changes bring about reliability improvements in later iterations of the parts.
The new way
Our innovative system virtually eliminates recurring set-up activity (see Fig. 1). The heart of the system is the software and holding fixtures. The software controls two vertical machining centers (VMCs), which perform the same basic tasks they would perform in a conventional machine shop. A three-axis VMC (the "plate machine") cuts a mix of parts from one raw aluminum plate (see Fig. 2). An operator then loads the parts, one at a time, into the five-axis VMC (the "edge machine"), which cuts the edges and flat surfaces as required until the part is ready for deburring.
In this system, the operator indicates the part numbers and quantities to be manufactured by pointing and clicking at a PC with a graphic user interface (see photo on p. 195). The system software automatically decides where to cut each part on the raw aluminum plate, even if the operator wants to produce several types of parts in the same run. A PC monitor displays a layout showing the proposed location of each part on each of three raw aluminum plates. The layout indicates what thickness of raw aluminum plate to place in each of three available positions. If the operator is satisfied with the layout, he or she can issue run and start commands from a PC.
The software reduces the total machining time by minimizing the number of tool changes. For example, it makes all cuts with Tool #1 on all the parts before switching to Tool #2 (unless doing so would violate some necessary sequence, such as drilling before tapping). The operator moves the parts from the plate machine to the edge machine, which offers the same intelligence as on the plate machine with respect to tool identification and tool placement. When the parts are removed from the edge machine, they are ready for deburring and finishing.
Challenges along the way
Our first challenge was to integrate the commercially available machine tools with our CAD/CAM software. To this end, we developed proprietary software in LabVIEW (National Instruments; Austin, TX) to provide a communication link between the milling-machine controllers and operator-attended PCs. An operator from a remote PC can command the milling machines to start machining specific parts in the desired quantities, spin the tool turret for tool inspection, or query the machine for its run states (on, off, waiting, and so forth). The milling centers can`t distinguish between commands sent through this channel (based on an RS422 communication protocol) and commands entered directly into their controllers.
We wanted the system to lay out the job automatically--in other words, to determine the work offset for each part on each of three raw aluminum plates. To solve this problem, we created a proprietary algorithm that takes into consideration the specifications for the individual parts, the 30-tool limitation of the turret, the effective lengths of each tool, and other parameters. It allows the system to cut several different types of parts in the same run. The system suggests intelligent tradeoffs; for example, to cut down on material waste, it might recommend running a 3/8-in. part on 1/2-in. plate that is not already full of parts rather than by itself on a 3/8-in. plate.
In a conventional system, a skilled machinist must make sure that the turret of the milling machine contains all the tools required for a particular job. To automate this task, we purchased the Kennametal Microlog Identification System, which consists of a code carrier, a read head, and a control interface. The code carrier, which mounts on an individual tool, consists of a single chip that stores a unique identification number and a coil connected to the chip. The read head supplies inductive energy to the code carrier and receives the identification number. The control interface processes the information, controls the data transmission, and interfaces with the host computer.
When the operator asks the system to run a job, the system takes inventory of the tools in the carousel to find out if all required tools are loaded. When the system finds the tools, it edits the milling machine`s instructions to indicate their actual locations. If a required tool is missing, the proprietary software directs the plate machine to present a slot with an unnecessary tool and requests the required tool in terms comprehensible to an unskilled operator--for example, it might say, "Install Tool #7," rather than "Install an end mill."
Because our layout algorithm allowed the three-axis milling machine to cut parts from plates of different thicknesses in the same run, we designed and built a quick-release, three-plate vacuum chuck. This chuck holds blocks of raw material in place by drawing a vacuum on the back side of them rather than applying a mechanical vise or clamp to the front side. It accommodates plates of different thicknesses and contains no hardware that might interfere with the path of the cutting tools.
The edge machine presented a more difficult problem with respect to material holding because we wanted it to cut as many as five surfaces of each part in one setup. To solve the problem, we designed and built a universal chuck that holds five jaws, at least one pair of which can hold virtually any part. We also developed a jaw-gripper mechanism to enable quick unlocking, repositioning, and relocking of the jaws.
To reduce the total machining time (and therefore the total cost) across different parts on the plate machine, we developed a proprietary algorithm that minimizes the number of tool changes. With this algorithm, the milling machine makes all cuts with Tool #1, for example, on all the parts before switching to Tool #2.
To eliminate the need for a skilled machinist as an operator, we developed a simple and intuitive graphical user interface. By pointing and clicking, the trained technician can select the part numbers and the quantities of the parts he or she wants to manufacture, then command the system to lay out the job, display the layout on the monitor and/or print a hard copy, and begin cutting parts. The technician can also enter or edit records in a parts and tools database.
Benefits for the optics industry
We have built a system in which milling machines can run an order with little intervention and with virtually no setup after the initial run. By automating most of the tasks that normally require the operator`s attention, we have freed him or her to focus on anomalies and to run other types of machines at the same time.
The system can run parts in any lot size required, even lot sizes of one, with little or no impact on the part cost (see Fig. 3). This capability allows a manufacturer of low-volume, high-mix products to establish flow lines that extend to the machining of the production parts. The system allows a factory to practice just-in-time (JIT) manufacturing and to demand flow techniques in their purest form to reduce the costs associated with maintaining a large inventory. Virtually any company could set up our system and begin producing parts with less investment in equipment and personnel than a conventional computer numerical control (CNC) shop would require. o
ACKNOWLEDGMENTS
New Focus engineer Dave Arnone was both inventor of this rapid-response, flexible manufacturing system and project manager during its development. Support for the project came from DARPA and from Focused Research.
Manufacturing engineer Ken Wallace (standing) and Rosie Abriam demonstrate user interface of the automated manufacturing system developed at New Focus subsidiary Focused Research.
FIGURE 1. With automated manufacturing system, unit cost of a typical prismatic part with six sides of work remains constant regardless of quantity produced. Typical costs for short runs and high-production runs are substantially higher for orders below 25 parts.
FIGURE 2. Lens mounts (left) and housing boxes, each on a metal plate about 12 in. square, are processed by the plate machine at the same time, demonstrating the system`s ability to produce mixed parts in the same small lot.
FIGURE 3. The small-lot capability of the automated manufacturing system makes it possible to turn out lens mounts of different sizes in the same production run.