Lean Improvement
SMED Quick Changeover: Break Changeover Losses into Work That Can Be Reorganized
The direct answer for SMED quick changeover is to break the changeover into observable work elements, distinguish internal work that must be completed while equipment is stopped from external work that can be completed earlier or later while it runs, and then shorten actual production occupancy through externalization, parallel work, simplification and standardization. The goal is not to rush people or reduce first-piece, safety or quality confirmation, but to eliminate waiting, searching, repeated adjustment and disorganized coordination.
Define the Changeover Start, End and Scope First
Changeover analysis must define which event starts the clock and which ends it. Common definitions involve the last good unit of the previous product, stable output of the next product or completion of first-piece approval. The plant should standardize the choice and disclose whether adjustment and first-piece stages are included.
Using only equipment restart as the endpoint can shift subsequent first-piece waiting and adjustment losses into OEE Quality or Performance. Recording only total shift-report time cannot distinguish preparation, removal and installation, adjustment and waiting. Retain the complete timeline and then map it to loss categories using consistent rules.
| Changeover stage | Typical work | Main loss risk | Data evidence |
|---|---|---|---|
| Preparation | Prepare documents, materials, tooling and people | Searching and confirmation begin only after shutdown | Material calls, issues, checks and readiness times |
| Shutdown and removal | Remove old tooling, materials or programs | Illogical sequence and unsuitable tools | Equipment status and work start/end times |
| Installation and setup | Set new tooling, recipe and parameters | Repeated adjustment and version errors | Version, scans and parameter confirmation |
| First piece and adjustment | Trial production, measurement, judgment and correction | Waiting for judgment, unclear standards and repeated trial adjustment | First-piece, inspection and adjustment events |
| Stable recovery | Reach the specified cycle time and quality condition | Low-speed running and hidden startup losses | Cycle, output and good-unit sequence |
Internal and External Work Depend on Current Technical Conditions
Internal work must be performed with the equipment stopped; external work can be completed while it is running. This classification is not inherent in the task name but determined by current equipment, tooling, safety and process conditions. The same work may move from internal to external after a modification.
Safety and quality constraints must be checked before externalization. Preassembling tooling, preparing materials early and prechecking programs can often be evaluated as external work, but hazardous energy, equipment protection or unauthorized parameter changes must never bypass established controls to save time.
| Work example | Current classification judgment | Improvement direction | Confirmation that cannot be omitted |
|---|---|---|---|
| Tooling and material preparation | Usually can be externalized | Kit and precheck against a checklist before changeover | Material, tooling version and condition |
| Program or recipe preparation | Depends on equipment and permission conditions | Pre-download or prevalidate controlled versions | Version, effectiveness and approval permissions |
| Old-tooling removal | Usually internal work | Quick connections, positioning and motion simplification | Shutdown and hazardous-energy control |
| New-tooling installation | Usually internal work | Standardized positioning and less adjustment | Mechanical, process and safety conditions |
| First-piece inspection | Judge according to the process boundary | Parallel information flow and early gauge preparation | Inspection requirements and release responsibility |
Use a Work-element Sheet to Record Facts, Not Just Total Elapsed Time
During observation, record each action in sequence, noting the performer, object, start and end, waiting, walking and dependencies. Video can be an approved internal analysis tool but is not required. Without suitable authorization, use multiple observers and time records, and do not display site footage in public materials.
Changeovers vary with product combination, people and equipment condition. Retain sample context and do not build a standard from the single smoothest run. The initial purpose of observation is to identify waiting and interface problems, not to evaluate individual speed.
- Record work: describe it with a verb and object, such as ‘verify tooling identity,’ rather than ‘prepare.’
- Record dependencies: identify the material, tool, information or other person's completion needed before continuing.
- Separate waiting: distinguish pure waiting from actual work and retain the reason for waiting.
- Tag conditions: record product combination, team, equipment condition and abnormalities for stratified comparison.
- Retain abnormalities: rework, wrong versions and repeated adjustments must not be edited into an idealized process.
Improvement Sequence: Externalize First, Then Parallelize and Simplify
The first round typically moves preparation, checks and closing work that do not require shutdown outside the internal window and establishes a kitting checklist. The second analyzes whether noninterfering tasks can run in parallel. Parallel work requires clear skills and communication signals; simply adding people can create mutual waiting.
Only the third round simplifies the remaining internal work technically and ergonomically—for example, through positioning, quick connections, fewer adjustment points and standard-parameter management. Every modification must be assessed for equipment, safety, quality and maintenance impacts; unapproved mechanical or control changes are not quick fixes.
| Improvement strategy | Applicable signal | Typical countermeasure | Validation focus |
|---|---|---|---|
| Externalization | Documents, materials or tools are still being sought after shutdown | Kitting, prechecking, preassembly and prevalidation | Is external preparation genuinely complete and correct? |
| Parallelization | One person waits for another, or serial tasks have no dependency | Role allocation and coordination signals | Are safety interference and responsibilities clear? |
| Simplification | Repeated fastening, adjustment, trial-and-error and walking | Positioning, quick connections and standardized workstations | Are quality and maintenance affected after modification? |
| Standardization | Sequence and results vary greatly between teams | Work-element breakdown, versions and training | Does it cover abnormalities and changeover combinations? |
| Earlier information flow | First-piece or recipe confirmation is delayed | Advance verification, interface notification and status transparency | Is approval responsibility still retained? |
Digitalization Validates Versions, Events and Standard Execution
MES or terminals can provide changeover tasks, product and process versions, kitting confirmation and work checklists. SCADA and the acquisition layer can record equipment stops, recovery, cycles and parameters. Quality systems or inspection records can provide first-piece judgments. A shared timeline makes every changeover stage reviewable.
Electronic checklists must not become a new entry burden. Information available automatically from work orders, scans and equipment events should not be entered again; people add only necessary confirmations and abnormalities. The system cannot automatically treat first-piece release or safety confirmation as complete; those responsibilities must follow plant procedures.
- Version error-proofing: verify the applicable relationships among current product, tooling, program and recipe.
- Kitting transparency: before internal changeover, show missing materials, tooling, gauges or approvals.
- Event records: retain objective times for stop, start, first piece, recovery and stabilization.
- Exception escalation: wrong versions, missing items and first-piece abnormalities enter Andon or the established response process.
- Effectiveness analysis: compare by product and team while observing both Performance and Quality losses.
Improve and Standardize One Changeover Combination First
Choose a frequent changeover combination with clear losses and well-defined process boundaries as the model. Establish a baseline and work-element breakdown, implement a small number of verifiable countermeasures, and update the standard only after validation across multiple changeovers under different conditions.
Results should report more than average changeover time. Also examine variation, first-piece quality, recovered cycle time, workload and safety execution. If average time falls while abnormalities rise, losses may have moved elsewhere or the standard may not yet be stable.
- Select the model: define product combination, equipment, boundaries and participating roles.
- Establish the baseline: break down internal, external, waiting, first-piece and recovery stages.
- Validate countermeasures: start with externalization and workflow improvement, then assess necessary technical modifications.
- Standardize: establish the kitting checklist, work sequence, roles and exception handling.
- Review continuously: stratify by changeover combination and observe time, quality, safety and recurring problems.
Shorten Unproductive Occupancy, Not Necessary Confirmation
SMED uses work-element breakdown, internal-to-external conversion, parallel work and simplification to reduce changeover downtime caused by waiting, searching and repeated adjustment. Clear start and end definitions and a complete timeline are prerequisites for determining whether improvement is real.
Digitalization can connect tasks, versions, kitting, equipment and first-piece evidence, but it cannot bypass safety or quality responsibilities. Use one changeover combination as a model, validate time, variation and startup quality together, and then standardize effective practices; this is more sustainable than pursuing faster motions alone.
Frequently asked questions
Does SMED mean every changeover must be reduced to a single-digit number of minutes?
No. SMED is a systematic improvement method. Specific targets depend on equipment, process, safety and quality conditions; the name should not be mechanically converted into one universal promise for every scenario.
Will adding changeover personnel always shorten the changeover?
Not necessarily. Only work that can run in parallel without interference may benefit. Too many people can also increase communication and space interference. Analyze work elements and dependencies before deciding how to parallelize.
Can first-piece approval be completed before shutdown?
A first piece normally needs to be produced under the new conditions before it can be approved, so this step cannot be omitted to compress internal time. Gauges, standards and judgment resources can be prepared in advance and information flow improved, but release responsibility remains unchanged.
Will a shorter changeover always improve OEE?
Changeover is an important source of availability loss, but final OEE also depends on Performance and Quality. If faster changeovers create first-piece defects or more low-speed adjustment, the overall effect may be offset, so all three factors must be checked.
