System Synthesis Inc.
An auto supplier had been manufacturing a 30-component device for several years and was fairly certain its assembly line held the optimum combination of manual labor and robots. But when the auto giants began pressing suppliers for lower costs, company engineers started looking for ways to economize. On a hunch that modest improvements might be gleaned from the tried-and-true assembly line, the supplier put it under the scrutiny of recently developed assembly system design software. Results were surprising.
The software suggested converting several manual stations to fixed automation and another to robotic operation. The manufacturer calculated nearly a 20% drop in assembly costs after making the conversions, and production capacity jumped by 5%.
At first glance, one might suspect the cost-saving package to be some version of design-for-assembly software. Not so. The idea behind DFA is to minimize the number of parts in a product while the assembly line package looks for the most efficient use of workstations. The packages complement each other. A third piece of software that is beyond the scope of this article looks at the order in which the pieces are assembled for even greater economy.
One goal of an assembly line is to perform as much work as possible for the lowest cost at each workstation. A direct software method finds solutions that satisfy the constraints. The usual method lets engineers guess at a solution to the system design and then analyze it to see how closely it comes to meeting technical and economic requirements. The general solution establishes the time available at a workstation for each worker and robot from
where T = time available at each workstation, sec; A = availability, a decimal portion of total time; S = number of shifts/day; Dy = work days/year; Py = yearly production; tmove = move time or in-out dead time, sec.; and tup = uptime, a decimal < 1.
The software then synthesizes the most cost-effective systems starting with availability = 1.0. For each subsequent system, the program uses a time figure slightly less than that required for the bottleneck station in the previous system (i.e., by reducing the availability) and calculates T. The process continues until additional stations are needed to perform any one task. This parallel-station tactic is usually called upon to alleviate a bottleneck. However, it’s rarely cost effective but may be necessary to satisfy production requirements.
Each task normally has more than one viable assembly method. Companies most often pick from manual labor, fixed automation, or programmable automation. Each of the methods or resources perform only some of the required tasks. For example, installing a heavy casting may best be done by fixed automation even though people with tools and robots could do it. Users must also prescribe economic constraints on the system such as minimum attractive rate of return, a capital recovery period, and production requirements.
The software then finds several usable solutions to the assembly system design problem. Each is defined by the maximum workstation time available. A method for rating each of the systems with user-alterable criteria readily pinpoints the best arrangement. Tables, graphs, and schematics provide specifications for the synthesized systems.
An interactive method in software simplifies establishing assembly cost and performance characteristics for each resource, the cost of additional hardware, and an estimate of the nominal time required for each applicable task. An experienced manufacturing engineer could assemble the information. This person would also know how to modify some of the data obtained from the software. For example, a particular robot or part presentation method may hold an advantage for performing some tasks but it might not be cost effective for more than one task. The engineer may substitute some specialized equipment as a necessary option. Manual labor should usually be an alternative because some tasks, especially in 3D assembly, are best performed by a human. Equipment vendors can provide hardware costs.
Parameters defined in this section of the software are based upon data derived from a wide variety of assembly systems. Cost and performance data for each workstation may not be exact because people often improve their efficiency over time, especially for a small production quantities. But the data is easily altered.
Resources also carry different costs and charges. For example, a manual workstation has a hardware cost for the bench, chair, and light. Hardware falls into cost categories for the resource and for the station which includes equipment (end effectors on robots, for instance) necessary for a resource to perform a single task. A fixed-automation station has no resource cost because each task generally requires a unique machine which possesses a station cost. And programmable automation, such as robots, have about four cost categories ranging from moderate (able to do a few tasks) to expensive (able to perform most tasks).
What makes finding the best assembly system difficult is that only a few arrangements of resources are applicable among thousands of possibilities. And finding the cost and performance characteristics for the resources may be a daunting chore.
The software normally uses one manual-labor category, one for fixed automation, and two or more for robots. Others occur when more labor classes are possible or when using specific devices such as ovens or automated pallet rotators.
An assembly plan for an air conditioning unit provides an example of what the software needs and what it produces. The table Software inputs shows a partial table of tasks and times for the case study. Tasks are identified by a single upper-case letter and each is color coded based upon the degree of difficulty. These tasks become the rows of the table and resources constitute the columns.
Users must also specify general limits on the time available at a workstation by considering:
• Work period in terms of days/ yr and the number of shifts available, usually 1, 2 or 3.
• Station-to-station move time in seconds indicates the workstation idle time.
• Units/pallet is almost always one, but can be used to prorate long in-out times, such as those usually needed for an automated guided vehicle.
• Production batch size describes the number of units to be produced in a work year.
• The maximum in-parallel stations is usually one, but a higher value could be required. This provides a limit to the number of solutions. The largest number of tasks at a workstation is usually determined by the available time but it could be controlled by a parameter called maximum tools at a station.
While it is a good idea to allow only consecutive tasks at a workstation, there are opportunities for what’s called “revisits” under certain conditions. Nonconsecutive task assignments can be beneficial, for example, when a person is required to work on either side of a robot’s work space, or before and after an oven or automated testing station.
The best system
Each system found using the process satisfies all technical and economic constraints. A rating system helps sort them out. The rating method lets the user assign weighting factors to unit cost, the investment required, and number of stations so that systems can be readily compared. Users may also assign targets for the unit cost, investment, and stations and let the program determine how well each system compares to those ideals.
Since the number of stations is usually important only when space is a premium, the first two factors often have higher weighting factors. The factors are expressed as percentages and must sum to 100. Examples appear in Ranking the systems. Each solution has a numerical rating which classifies the systems in descending order. The highest ranked solution is best. Users can see how each system scored and get an idea what adjustments might be made to the cost, station requirements, or the task-resource data to improve it.
The Availability factor is a fraction of the work year used to determine the upper limit on time available at any workstation. Actual unit cost is the sum of the investment capital recovery, labor charged for full shifts, and system operating and maintenance costs divided by the yearly production volume. Total investment includes the cost of hardware, engineering, design, installation, and debugging. Resources used shows the number of different workstations in the system. The asterisk in one of the resource columns indicates the type of bottleneck station.
This rating table is often the most valuable information produced by the software. Although better solutions tend to occur near the high end of a work shift for systems which contain direct labor, this is not necessarily the case for mostly automated systems. Users decide when the best solution should be used. For example, the table shows the best solution has at least one manual station as its bottleneck. That may be unacceptable. When that is the case, three of the top four systems in the table would have to be ignored. The software produces a variety of additional tables that detail a wide range of cost and performance information.
The layout illustrated in How to read an assembly schematic is one of the more interesting. It shows the task time and identifier, and a difficulty level for each of the three resources.
The schematic shows that Station 2 (MNL-2) and 5 (MNL-3) are bottlenecks and that Station 10 (P50-2) has nearly the same expected cycle time. Further study reveals that Station MNL-2 has at least one complex task while MNL-3 has at least one moderate task. Therefore, putting the best worker at MNL-2 might alleviate additional difficulty at the bottleneck. The two most difficult task stations in this system require about 81% of the system cycle time. Those task times could be increased about 20% (for the prescribed production volume, or less), if necessary.
The software also establishes a sensitivity of cost for variations in production volume. For example, the illustration Batch cost vs. yearly production shows the rate of change (slope of curve) in unit cost at the prescribed production volume as well as the system utilization. When the production volume differs by more than 5%, the whole system design process should be repeated using the new value. Most manufacturing systems are best for only a limited range (plus or minus a few percent) of production volumes.