SMED

Goal

SMED shortens the changeover of a machine, a line or a cell that switches from one part number to another. The setters make the first saving by reorganising their work, with no investment.

A long changeover pushes the shop to run large batches to absorb it. Large batches fill the stores and lengthen the time between order and delivery. A short changeover makes small batches pay, and the shop then produces at the pace of demand. For Shingo, producing in small lots with little stock presupposes short changeovers.

The deliverable is a changeover analysis sheet. It lists each operation with its duration, its type (internal or external) and the measure that moves it out of downtime or shortens it. A total line gives the changeover time after each stage. The project ends with a written changeover procedure that every shift applies.

Value stream mapping records the changeover time in the data box of each process step and shows where that time sets the batch size. SMED then works on one designated changeover, on one machine.

Usage

When to use it

• A bottleneck machine that changes over often: every minute of stoppage there takes a minute of capacity from the whole flow.
• Runs lengthened to absorb the changeover: shorten the changeover before cutting batch sizes and stock.
• A growing number of part numbers: more variants for the same volume mean more changeovers per week.
• Changeover time that varies from one setter to another: the time study shows the operations done from memory, which the written procedure then standardises.
• A planned shutdown outside the factory (a go-live, a system cutover, the refurbishment of a room): prepare during normal operation everything the shutdown does not require.

When not to use it

• Continuous production or a machine dedicated to a single part number: there is no changeover to reduce; look for the losses in the flow with value stream mapping.
• Stoppages caused mainly by breakdowns: the lost time comes from failures; look for their causes with a root cause analysis.

Description

Changeover time

The Lean Enterprise Institute (LEI) defines changeover time as the time elapsed between the last part of the finished run and the first good part of the next run. Trial parts, dimension corrections and the wait for inspection therefore belong to the changeover, even when the machine is already running. A time study that stops at restart misses the trials and the adjustments, which are often the longest part.

Internal and external operations

An internal operation can only be done with the machine stopped: mounting the new die, changing the collet on a lathe. An external operation can be done while the machine runs the previous or the next batch: bringing the die to the press, putting away the tooling of the finished run. Trial runs and adjustments also count as internal: the machine runs, but it does not yet produce a good part.

The sorting question asks what the operation requires of the machine, whatever the current practice. In the starting state, which Shingo calls the preliminary stage, nobody makes the distinction. The setter stops the machine, then fetches the tools, fills in the documents for the next run and waits for the forklift driver. The machine waits with him.

Shingo's three stages

After the preliminary stage, Shingo describes three stages. Each asks its own question of the same list of operations.

Stage 1 costs the least and returns the most. Shingo considers it the most important: sorting the operations and doing the external ones outside downtime usually cuts changeover time by 30 to 50%. The team buys nothing and leaves the machine as it is. The setter prepares before downtime and puts away after it.

Stage 2 revisits the classification of Stage 1. An operation is often internal because the tooling makes it so: a die that reaches its temperature only on the press, a tool adjusted in place for want of a common reference. Changing the tooling or the preparation makes the operation external.

Stage 3 shortens each remaining operation. Fastenings and adjustments take the most time at this stage. Shingo observes that a bolt takes many turns although only the last one tightens and only the first one loosens; the others only move the nut. An adjustment corrects a position that the tooling did not guarantee. A stop or a recorded value fixes that position, and the adjustment disappears. Two setters working in parallel on either side of the machine eliminate their trips back and forth, provided they follow a written division of tasks.

Shingo puts mechanisation last. A company that mechanises an operation it could have eliminated or moved out of downtime pays for equipment to speed up work it could have removed from the changeover.

The stages overlap where needed: a project can shorten an internal operation before the sorting is finished, and each stage produces its own gain.

Running a SMED project

1. Choose the changeover. One machine and one switch from a part number to another: the most frequent one or the one on the bottleneck machine. Bring together the setters from each shift, a production manager and, if the tooling has to change, maintenance or the industrial engineering office.
2. Film a real changeover. From the last part of the run to the first good part, on an ordinary day, done by its usual setter. The team reviews the film with the setter, who comments on each move. Filming staff touches on employment law. Art. 26 OLT 3 prohibits surveillance or monitoring systems intended to monitor the behaviour of workers at their workplace (para. 1). A system needed for another reason must be designed and placed so that it harms neither their health nor their freedom of movement (para. 2). The camera therefore frames the machine and the tooling. The production manager informs the team before filming. Our fiche on observation explains how to observe work on the shop floor.
3. Break down and time. One line per operation, with its duration. Measure the total changeover time first, then the time of each operation.
4. Classify each operation as internal or external, by what it requires of the machine: Stage 1.
5. Convert. Go back over each internal operation and look for what would make it external: Stage 2.
6. Shorten the internal, then the external. The LEI treats the two in that order. Internal time costs lost production, external time costs setter time.
7. Standardise. Write the procedure, the checklist and the division of tasks before, during and after downtime. Train each shift. Time the following changeovers to check that the time holds.

Pitfalls

Buying before sorting. A project that starts with hydraulic clamping or a motorised cart skips the stage that returns the most for the lowest price.

Leaving external work without an owner. An operation moved out of downtime is still work. If nobody plans it, it comes back: the setter arrives, the tool is not ready, the machine waits. The procedure states when preparation starts and who does it.

Timing a staged changeover. Prepared for the camera, the changeover no longer shows the searching and the waiting that Stage 1 moves out of downtime. The team films an ordinary day and announces that it will use the film to analyse the operations, not to assess the setter.

Treating adjustments as craft. Setters see trials and corrections as the heart of their trade and leave them out of the changeover. The time study counts them up to the first good part, and Stage 3 treats them as a positioning defect.

Not deciding what the time gained is for. The freed minutes go to capacity or to shorter runs. If nobody chooses, batch sizes stay where they were and stock does not fall.

Not writing the procedure down. Without a written procedure and checklist, each shift goes back to its habits and the time creeps back up.

AI considerations

A vision model cuts the video of a changeover into time-stamped operations and supplies a draft time study. The setter corrects the cuts, because only he knows where an operation ends. A language model receives the list of operations, asks the Stage 2 question of each one and suggests the known means: presetting, jigs, quick clamping. Its suggestions form the agenda of the conversion session.

The model does not know what the list leaves out: the tolerances of the part, the condition of the tooling. An operation that is internal for a safety reason, such as a locked-out machine or an open guard, stays internal whatever gain the model claims. The team settles the classification at the machine, with the setter and the person responsible for safety.

A video in which the setters can be recognised contains personal data within the meaning of the Federal Act on Data Protection (FADP, art. 5). Sending it to an external vision model entrusts its processing to a processor (art. 9). If that provider processes the data outside Switzerland, the rules on disclosure abroad apply (art. 16). A model run in-house keeps the video on site. Blurring faces before sending reduces what identifies the setters.

Examples

Analysis sheet for a changeover on a bar-turning machine, a Swiss-type automatic lathe with a sliding headstock, switching from a steel watch screw (run A) to a brass shaft (run B). Each column gives the internal time after one stage, which is the changeover time.

Stage 1 takes 40 minutes off the changeover, 31% of the starting time, without shortening any operation that stayed internal: rows 1, 7a and 8 move out of it. The final 31 minutes remain above the ten-minute target, and the sheet shows where the project goes next: the 8 minutes of trials in row 6, then rows 2 and 4, at 6 minutes each.

Visualisations

The analysis sheet is a table: one row per operation, one column of internal time per stage, a total at the foot of each column. The stage diagram places the same changeover on a time axis. A frame marks the changeover, from the last part A to the first good part B. The internal operations sit inside it, the external ones before or after. From one timeline to the next, the reader sees the frame shrink. The operation numbers are those of the sheet.

Cost

Tooling

A mobile phone films the changeover, and a video player that displays elapsed time is enough to cut it into operations. A spreadsheet holds the analysis sheet and recalculates the changeover time at each reclassification. In the workshop, sticky notes in two columns, internal and external, make the sorting visible to the whole team. The means in the procedure are physical: a laminated checklist at the machine, a changeover cart prepared before downtime, a tool presetting bench. The spaghetti diagram from workplace layout analysis, drawn for the setter during the changeover, shows the trips to the stores that Stage 1 eliminates.

Sources

• Shigeo Shingo, A Revolution in Manufacturing: The SMED System, trans. Andrew P. Dillon, Productivity Press, 1985: the method, the distinction between internal and external operations, the preliminary stage and the three stages, the means of each stage, the 30 to 50% gain of Stage 1 and the link between changeover time and small-lot production.
• Lean Enterprise Institute, Single Minute Exchange of Die, Lean Lexicon: the target of a changeover time under ten minutes and Shingo's two contributions, separation and conversion.
• Lean Enterprise Institute, Changeover, Lean Lexicon: the definition of a changeover and the measurement of its time, from the last part of the finished run to the first good part of the next.
• Lean Enterprise Institute, Setup Reduction, Lean Lexicon: the six steps of setup reduction, from measurement to standardisation.
• Swiss Confederation, Ordinance 3 to the Employment Act (OLT 3 / ArGV 3) (Health Protection), SR 822.113, art. 26: the prohibition of surveillance or monitoring systems intended to monitor the behaviour of workers at their workplace (para. 1) and the conditions set for a system needed for another reason (para. 2).
• Swiss Confederation, Federal Act on Data Protection (FADP), SR 235.1: the definition of personal data (art. 5), processing by processors (art. 9) and the disclosure of personal data abroad (art. 16).