In this four-part series, we take an in-depth look at how to design an effective work environment. Part one discusses the elements of continuous-flow work cells. Part two considers how to enhance the efficiency of such work cells. Part three explores the 5S methodology. In this, the last part of the series, we look at single-minute exchange of die (SMED).   

Changing from producing one part to another is an area worth special attention. This topic, like many in lean Six Sigma, is a big one. Many books have been written on the subject, so in this article, I’ll present a brief overview of the topic.

Single-minute exchange of die (SMED) is one of the many lean production methods for reducing waste in a manufacturing process.¹ It provides a rapid and efficient way of converting a manufacturing process from running the current product to running the next product. This rapid changeover is key to reducing production lot sizes and thereby improving flow (mura). The SMED concept was credited to Shigeo Shingo, one of the main contributors, along with Taiichi Ohno, to the consolidation of the Toyota Production System.

The term “single minute” doesn’t mean that all changeovers and startups should take only one minute, but that they should take less than 10 minutes (in other words, “single-digit minute”).² Closely associated is a yet more difficult concept, one-touch exchange of die, (OTED), which says changeovers can and should take less than 100 seconds.

Reverse-engineering economic lot size

Table 1 shows the increase in effective operating time caused by a production changeover. SMED is the key to reducing operating time and gaining manufacturing flexibility.

Table 1: Changeover time vs. operation time

The “economic lot size” (or EOQ) is a well-known, and heavily debated, manufacturing concept.³ Historically, the overhead costs of retooling a process were minimized by maximizing the number of items that the process should construct before changing to another model. This makes the changeover overhead per manufactured unit low. According to some sources, optimum lot size occurs when the interest costs of storing the lot size of items equals the value lost when the production line is shut down.

The difference, for Toyota, was that the economic lot size calculation included high overhead costs. Land costs in Japan are very high, and therefore it was very expensive for Toyota to store its vehicles. The result was that its costs were higher than other producers because it had to produce vehicles in uneconomic lots. Engineer Shingo could do nothing about the interest costs, but he had total control of the factory processes. If the changeover costs could be reduced, then the economic lot size could be reduced, directly reducing expenses. Indeed, the whole debate concerning economic lot size becomes restructured, if still relevant. It should also be noted that large lot sizes require higher stock levels to be kept in the rest of the process, and these—more hidden costs—are also reduced by the smaller lot sizes made possible by SMED.

Throughout a period of several years, Toyota reworked factory fixtures and vehicle components to maximize its common parts, minimize and standardize assembly tools and steps, and utilize common tooling. These common parts or tooling reduced changeover time. Wherever the tooling could not be common, steps were taken to make the tooling quick to change.

Fine-tuning the die changeover

Toyota found that the most difficult tools to change were dies on the large transfer-stamping machines that produce car bodies. The dies—which must be changed for each model—weigh many tons and must be assembled in the stamping machines with tolerances of less than a millimeter; otherwise, the stamped metal will wrinkle, if not melt, under the intense heat and pressure.

When Toyota engineers examined the changeover, they discovered that the established procedure was to stop the line, let down the dies by an overhead crane, position the dies in the machine by human eyesight, and then adjust their position with crowbars while making individual test stampings. The existing process took from 12 hours to almost three days to complete.

Toyota’s first improvement was to place precision measurement devices on the transfer stamping machines, and record the necessary measurements for each model’s die. Installing the die against these measurements, rather than by human eyesight, immediately cut the changeover to a mere 90 minutes.

Further observations led to further improvements: scheduling the die changes in a standard sequence as a new model moved through the factory, dedicating tools to the die-change process so that all needed tools were nearby, and scheduling use of the overhead cranes so that the new die would be waiting as the old die was removed. Using these processes, Toyota engineers cut the changeover time to less than 10 minutes per die, and thereby reduced the economic lot size below one vehicle.

The success of this program contributed directly to just-in-time manufacturing, which is part of the Toyota Production System. SMED makes load balancing much more achievable by reducing economic lot size and thus stock levels.

Benefits of implementation

Shingo, who created the SMED approach, claims that in his data collected between 1975 and 1985, average setup times have reduced to 2.5 percent of the time originally required, an improvement of 40 times the original rate.⁴

However, the power of SMED is that it has a lot of other positive effects that come from systematically looking at operations. These include:

• Stockless production, which drives capital turnover rates
• Reduction in footprint of processes, including reduced inventory that frees up floor space
• Productivity increases or reduced production time
• Increased machine work rates from reduced setup times, even if number of changeovers increases
• Elimination of setup errors and trial runs, which reduces defect rates
• Improved quality from fully regulated operating conditions in advance
• Increased safety from simpler setups
• Simplified housekeeping from fewer tools and better organization
• Lower expense of setups
• Operator preferred because easier to achieve
• Lower skill requirements because changes are now designed into the process rather than a matter of skilled judgment
• Elimination of unusable stock from model changeovers and demand estimate errors
• Goods are not lost through deterioration
• Ability to mix production allows flexibility and further inventory reductions, also revolutionized production methods (large orders ≠ large production lot sizes)
• New attitudes about controllable work process amongst staff

Implementation techniques

Shingo recognizes eight techniques that should be considered when implementing SMED.⁵

1. Separate internal setup operations from external
2. Convert internal to external setup
3. Standardize function, not shape
4. Use functional clamps or eliminate fasteners altogether
5. Use intermediate jigs
6. Adopt parallel operations (see figure 1)
7. Eliminate adjustments
8. Mechanization

External setup can be done without the line being stopped, whereas internal setup needs the line to be stopped.

Shingo suggests that SMED improvement should pass through four conceptual stages:⁶

1. Ensure that external setup actions are performed while the machine is still running
2. Separate external and internal setup actions to ensure the parts all function
3. Implement efficient ways of transporting the die and other parts
4. Convert internal setup actions to external
5. Improve all setup actions

Seven steps to SMED

There are seven basic steps to reducing changeover using the SMED system (see figure 1):

1. Observe the current methodology (A)
2. Separate internal and external activities (B). Internal activities are those that can only be performed when the process is stopped, while external activities can be done while the last batch is being produced, or once the next batch has started. For example, go and get the required tools for the job before the machine stops.
3. Convert, where possible, internal activities into external ones (C) (preheating tools is a good example).
4. Streamline the remaining internal activities by simplifying them (D). Focus on fixings—Shingo rightly observed that it’s only the last turn of a bolt that tightens it—the rest is just movement.
5. Streamline the external activities so they are of a similar scale to the internal ones (D).
6. Document the new procedure and actions that are yet to be completed.
7. Do it all again: For each iteration of the above process, a 45-percent improvement in setup times should be expected, so it may take several iterations to cross the 10-minute line.

Figure 1: SMED method illustrated

Figure 2 shows four successive runs with learning from each run and improvements applied before the next.

• Run 1 illustrates the original situation.
• Run 2 shows what would happen if more changeovers were included.
• Run 3 shows the effect of the improvements in changeover times that come from doing more of them and building learning into their execution.
• Run 4 shows how these improvements can get you back to the same production time but now with more flexibility in production capacity.
• Run N (not illustrated) would have changeovers that take 1.5 minutes (a 97-percent reduction) and whole shift time reduced from 420 minutes to 368 minutes a productivity improvement of 12 percent.

Figure 2: Four successive runs using SMED

Table 2: Proportion of time for various operations

Look for:

• Shortages, mistakes
• Inadequate verification of equipment, which causes delays. This can be avoided by checking tables, especially visual ones, and doing setup on an intermediary jig.
• Inadequate or incomplete repairs to equipment, causing rework and delays
• Optimization for least work as opposed to least delay
• Unheated molds, which require several wasted “tests” before they will be at the temperature to work
• Using slow, precise adjustment equipment for the large, coarse part of adjustment
• Lack of visual lines or benchmarks for part placement on the equipment
• Forcing a changeover between different raw materials when a continuous feed, or near equivalent, is possible
• Lack of functional standardization, i.e., standardization of only the parts necessary for setup, e.g., all bolts use same size spanner, die grip points are in the same place on all dies
• Too much operator movement around the equipment during setup
• More attachment points than actually required for the forces to be constrained
• Attachment points that take more than one turn to fasten
• Any adjustments after initial setup
• Any use of experts during setup
• Any adjustments of assisting tools such as guides or switches

Record all necessary data using a form such as that shown in figure 3.

Figure 3: SMED data recording form

Parallel operations using multiple operators

By taking the “actual” operations and making them into a network, which contains the dependencies, it is possible to optimize task attribution and further optimize setup time. Issues of effective communication among the operators must be managed so that safety is ensured where potentially noisy or visually obstructive conditions occur. This idea is illustrated in figure 4.

Figure 4: SMED with multiple operators

[1] The material on SMED is from a particularly good article on Wikipedia at http://en.wikipedia.org/wiki/Single-Minute_Exchange_of_Die.
[2] Shingo, Shigeo. A Study of Toyota Production System: From an Industrial Engineering Viewpoint. (Productivity Press, 1989.) p. 70.
[3] Goldratt, Eliyahu. Theory of Constraints. (North River Press, 1990.) p. 40.
[4] Shingo, Shigeo. A Revolution in Manufacturing: The SMED System. (Productivity Press, 1985.) p. 113.
[5] Shingo, Shigeo. A Study of Toyota Production System: From an Industrial Engineering Viewpoint. (Productivity Press, 1989.) p. 47.
[6] Shingo, Shigeo. A Revolution in Manufacturing: The SMED System. (Productivity Press, 1985.) p. 27.