溶接セルの稼働率と切換え：両方をどのように分析するか

Every capacity case for a welding cell contains one optimistic number. It is usually arc time, expressed as a percentage and calculated by dividing the weld time on a sample part by an eight-hour shift. The cell then arrives and settles into a utilization figure twenty or thirty points below the case. Nothing broke; the missing points were never in the calculation – changeover, loading that waits on a forklift, the first-part check, and the shift change in the middle of it all.

Utilization and changeover are the same subject from opposite ends. Utilization is how much of the available time the cell welds; changeover is the largest single claim on the time it does not. Analysing both before the purchase order turns a hope into a number that survives a production schedule.

Defining Utilization Before Measuring It

The first discipline is agreeing what the denominator is, because utilization arguments almost always turn out to be denominator arguments. Four definitions, all legitimate, all producing different numbers:

• Calendar time: all hours in the period. The harshest basis and the only one supporting a straightforward financial case.
• Shift time: scheduled production hours, excluding nights and weekends. The most common basis in single or double shift plants.
• Planned production time: shift time less planned maintenance, breaks and training. The fairest basis for judging the equipment rather than the schedule.
• Loaded time: hours when parts and an operator were actually available. The most flattering basis, and the one that hides material supply problems.

Whichever basis is chosen, write it beside the number and keep using it. The remote monitoring and OEE reporting article applies the same discipline: OEE factors are comparable only when their definitions are fixed. Capacity utilization is the standard framing, and its lesson is that the headline rate means nothing without the reference capacity stated beside it.

Where the Missing Hours Actually Go

Once the denominator is fixed, the losses can be listed. On a typical multi-variant cell they fall into a small number of buckets, and most plants can fill them in from memory within a few percent:

• 取り替え: fixture swap, programme selection and first-part verification. On high-mix cells this is routinely the largest single loss.
• Load and unload: time when the cell is complete and waiting for a part, or for an operator who is elsewhere.
• First-part inspection: necessary and frequently unbudgeted. Where inspection happens away from the cell, the cell waits.
• Unplanned stops: nozzle changes, window cleaning, wire snarls, sensor faults and the small adjustments operators make.
• Programming and teaching: new part introduction consumes cell time where teaching happens in place rather than offline.

Programming deserves particular attention, because teaching a new part on the cell competes with production for the same hours. The programming cost per programme analysis therefore belongs inside the utilization discussion; offline programming moves that cost off the critical path.

Analysing Changeover Properly

Changeover is measured in two halves, and the distinction drives every improvement. Internal time is time when the cell cannot weld. External time is preparation done while the cell is still running the previous batch. The majority of changeover reduction available on a welding cell comes from converting internal tasks into external ones, not from doing the internal tasks faster:

• Pre-staged fixtures: the next fixture built and verified on a bench while the current batch runs.
• Preset parameters: programmes and wire settings stored by part number rather than adjusted at the cell.
• Standardised pallet interface: one mechanical and pneumatic interface means every fixture locates the same way, removing the shimming that quietly consumes internal minutes.
• Consumable readiness: wire type, gas and nozzles confirmed before the changeover starts, not discovered halfway through.

Single-minute exchange of die is the established method, and its central move – separate internal from external, then convert – applies to a welding cell almost unchanged. The mechanical enablers are covered in the quick-change and pallet systems article, and the design choices behind repeatable changeover in the fixture design guide. Batch size then decides how much any of this matters: doubling the batch halves changeover cost per part, which is a scheduling decision worth modelling before blaming the machine.

Building a Capacity Case That Survives

A defensible case multiplies rather than divides. Start from available hours, subtract changeover across the expected number of changeovers, subtract load and unload time per part, divide by the measured cycle time, and apply a quality yield:

• Use measured cycle times: from a demonstration run at production rate on production-representative parts, not a theoretical weld speed.
• Model the real mix: four part numbers in unequal volumes produce four different utilization outcomes, weighted by actual demand.
• Leave headroom: a case that needs ninety percent utilization to pay back fails the first time something goes wrong.

Two follow-on checks make the case durable. Ask what the cell does at low utilization, since most plants buy for peak demand and live at average demand; the cycle time optimization work is where recovered minutes appear. Then plan the expansion path, because a second cell sharing an operator changes the arithmetic entirely – the multi-cell expansion planning article sets out the shared-services version. 総機器効率（OEE） remains the cleanest summary, provided its three factors are reported separately under the definitions agreed at the start.

よくある質問