Alternatives for the Future of Machine Cells
The functions and equipment for the cell described above are by no means exclusive. Various alternative configurations can be used to increase flexibility -or productivity. For example, the machining cell becomes considerably more versatile if a four or five axis machining center is used in place of a vertical axis CNC mill. The much greater expense of a machining center is not justified for machining parts from the family typified by figure 3-1 but it would allow the cell to tackle complex parts that require too many set-ups with a vertical axis mill or that require contour cutting. As a second example, the lathe could be replaced with a multi- spindle chucker which would considerably increase the number of turned parts per hour. For many small- batch precision parts, however, high production volumes are not an issue.
The way in which the cell locates and then acquires new parts is flexible, but slow. The vision system looks at each part, determines the coordinates of the part and sends them to the cell host. The cell host then sends them to the mill-loading robot which carefully grips the parts and transfers them, one by one, to one of the machine tools. Afaster approach is to use a pallet with fixtures to hold the parts so that they arrive at the cell with a known orientation. The pallet and cart may be aligned with mechanical locating devices, or the pallet may be located using the vision system -just once. This approach makes less use of the vision system, and is less flexible since it presumes that parts are accurately located on the pallet when they arrive. However, if the parts have arrived from another cell they will have been put down in a systematic and accurate fashion. The only other requirement is to design the pallet so that the parts do not jostle while they are transported. The vision system is used only to count the parts, and check that they match the expected description. Even if the vision system is used to establish the position and orientation of parts, it is still advantageous to begin with a known approximate part orientation. This reduces the likelihood of ambiguous part orientations and can improve the speed and accuracy of the solution.
The sensors used in the cell could include pressure sensors, force sensors, vibration sensors, and optical measurement equipment. These devices would allow the cell to check for problems and do some rudimentary troubleshooting. For example, proportional sensors such as LVDTs or pneumatic gages on the fixtures could display not only whether parts were aligned but how well they were aligned compared to previous parts in the batch. Once random noise is filtered out of such measurements, they become very useful in indicating trends such as a gradual deterioration in robot accuracy or fixture performance. The goal is to discover such trends before they become a serious problem.
Sensors can also be used to monitor the processes within the cell in an effort to optimize them. For example, in the machining process one might emulate some of the senses that an experienced machinist uses when running his machine. Thus, in an attempt to automate a machinist's use of tactile feedback, accelerometers could be mounted in strategic locations on the machine tool to measure the vibration of the system composed of the machine tool, fixture and work piece. This information would be used to identify speeds. feeds, and fixture configurations which maximize chip removal while minimizing vibration. While a machinist might judge the rate of power consumption of a machine tool by listening to the spindle drive motor, a volt meter in combination with an ammeter could measure the drive motor's power consumption.
Checking work piece dimensions, a simple matter of reading a value from a caliper or micrometer for a machinist, might be accomplished with displacement transducers mounted in specially designed tool holders. This, however, has the drawback of being limited by the accuracy of the machine tool itself and occupies the machine with measurement tasks when it could be cutting metal. The list can be extended to increasingly obscure parameters that are normally monitored by a machinist operating a milling machine but until better theoretical models of the machining process are available there is little point in monitoring any but the most basic variables.
The devices mentioned above are more delicate than the simple sensors specified in section 4.3 and they often require expensive signal processing electronics; but the real reason for not including them in the first implementation of the flexible cell is that they require sophisticated software support for control and diagnosis. A considerable amount of research and development would be required to produce such software.
Most of the alternatives to the near-term cell, including those mentioned above in this section, would increase the first cost of the cell. There are very few alternatives that will decrease the cost of the cell because a minimum cell configuration has been selected. The one exception to this rule is to wait. We can wait until machine tools become more flexible, until controllers become more powerful and versatile, until the software required for flexible cells is less novel, and then the cost of the cell will go down. Unfortunately if we do wait, we are no longer building the "Factory of the Future." Also, many of the development costs pertain to the first cost of the first cell. Once a couple of cells have been built, the development costs become minor.
