Question 3 of 8: SMED and Kanban in Just-in-Time Production

Question 3: SMED and Kanban in Just-in-Time Production (20 marks)

Question text not reproduced: the examination questions are © Engineers and Geoscientists BC. Open the official past paper (linked at the top of this page) to read the question, then follow the worked solution below.

(a) SMED: Definition, Mechanism, and an Example

SMED (Single-Minute Exchange of Die) is a lean methodology, developed by Shigeo Shingo at Toyota, for reducing equipment changeover time — the time between the last good part of one production run and the first good part of the next — down to “single-digit minutes” (under ten). It works by separating every changeover task into two categories and then systematically shrinking both:

1. Separate internal from external setup. Internal setup elements can only be done while the machine is stopped (removing the current die, bolting in the new one); external elements can be done while the machine is still running the previous job (fetching the next die and its tools, pre-checking bolt torque specs, staging fixtures on a cart beside the machine). The first and single biggest gain comes simply from correctly sorting existing practice into these two categories — most plants, before SMED, perform many genuinely external tasks while the machine sits idle purely out of habit.
2. Convert internal setup to external wherever possible. Pre-heating a die, pre-assembling a sub-fixture off-line, or standardizing bolt/clamp sizes so a single wrench serves every die are all conversions that move work from the stopped-machine column to the running-machine column, directly cutting downtime without touching the remaining internal steps at all.
3. Streamline both the remaining internal and external elements. Replace bolts with quarter-turn clamps or cams, use guide pins/locating stops so alignment is automatic rather than measured by hand, and use standardized die heights so no shimming is needed — each of these shrinks the time the remaining (now-minimized) internal steps actually take.

Example. A stamping press changing dies between two part numbers originally takes 90 minutes: the operator waits for the crane to bring the next die (internal, but purely a waiting/logistics problem), unbolts eight mounting bolts of varying sizes, manually shims the new die to the correct height, and re-torques all eight bolts. Applying SMED: the next die is pre-staged on a cart beside the press and inspected while the current die is still running (external); the eight assorted bolts are replaced with quarter-turn clamps on a standardized die-base height (streamlining); and a locating pin fixes horizontal alignment automatically (eliminating the manual shim step). The changeover, now only “crane the pre-staged die in, engage four clamps, verify with a go/no-go gauge,” falls from 90 minutes to under 8 — enabling smaller batches (since setup no longer dominates the cost trade-off), which is precisely the enabler JIT/kanban production depends on.

(b) A Situation Where Kanban Would Not Be Advisable

Kanban is a pull-based visual signal system: a downstream station only produces (or a supplier only ships) when an empty kanban card or bin arrives, and the number of cards in circulation for a part sets a hard ceiling on the WIP for that part. Sizing that card count correctly assumes demand for the part is reasonably stable and repetitive, so a fixed number of cards, once tuned, keeps the line fed without either starving or over-producing.

Example situation: a job shop building highly customized, low-volume, engineer-to-order equipment — e.g., a manufacturer of one-off industrial test rigs, where each order has a different bill of materials, different routing through the shop, and no two consecutive jobs are alike. Kanban is not advisable here because:

• There is no repeating part number to attach a card to — kanban controls the replenishment of a specific, recurring item between two fixed points in a process; a part that is built once and never again has no steady-state demand rate for a card count to be tuned against.
• Demand and routing are both highly variable and unpredictable job-to-job, so any fixed number of cards is wrong almost immediately: too few cards starves the (constantly-changing) bottleneck resource, too many lets WIP balloon for parts that may never be ordered again.
• Kanban's core benefit — a simple, decentralized visual signal that avoids the overhead of a full scheduling system — is lost when routings differ every job, because there is no stable flow for the visual signal to regulate; a project-scheduling or MRP-based push system, which explicitly plans each unique job's due date and material requirements, fits this environment far better.

By contrast, kanban is well-suited to the opposite case: a repetitive, high-volume assembly line consuming the same fasteners, sub-assemblies, or components shift after shift, where a stable average consumption rate makes a fixed card count both calculable and effective.