Takt Time, the Real Constraint and a Misplaced SMED Event
A pump assembly value stream has demand of 18,400 units per month over 20 working days, two shifts of 450 available minutes each. Current state: CNC machining 42 s cycle with three changeovers per day of 90 minutes each; welding 61 s cycle on a single machine, 92% uptime and 4% rework that must be re-welded; a wash step that processes batches of 300; assembly 55 s; final test 48 s. The plant manager wants to meet demand next quarter without capital spend. A kaizen team has proposed a three-day SMED event on the CNC to cut changeover from 90 to 30 minutes, and has already booked the facilitators. The welder currently runs through breaks to keep up, and CNC operators boast about never letting the machine idle.
1. Calculate takt time and the effective capacity of the welding operation. Which operation is the constraint, and how big is the gap?
Read the worked answer
Available time: 2 × 450 × 60 = 54,000 s/day → × 20 days = 1,080,000 s/month.
Takt time = 1,080,000 / 18,400 = 58.7 s per unit.
Weld effective capacity must account for uptime and rework, not just cycle time:
- Rework multiplier: 61 × 1.04 = 63.4 s of welder time per good unit.
- Uptime: 63.4 / 0.92 = 68.9 s of calendar time per good unit.
Capacity = 1,080,000 / 68.9 ≈ 15,700 units/month - roughly 2,700 units short (15%) of the 18,400 requirement.
Required effective cycle to hit demand: 1,080,000 / 18,400 = 58.7 s of calendar time, i.e. 58.7 × 0.92 = 54.0 s of welder time per good unit, so rework-adjusted cycle must fall to 54.0 / 1.04 ≈ 52 s - a 15% improvement from 61 s.
Conclusion: welding is the constraint and 61 s already exceeds takt before uptime and rework are applied. Any improvement elsewhere cannot raise throughput until weld is relieved.
2. Evaluate the SMED proposal. Is it the right first move, and what is the correct sequence of actions?
Read the worked answer
CNC arithmetic: changeover time is 3 × 90 = 270 min = 16,200 s of the 54,000 s available - 30% of capacity. Capacity = (54,000 − 16,200) / 42 = 900 units/day = 18,000 units/month, also below the 18,400 requirement. With 30-minute changeovers: (54,000 − 9,000) / 42 = 1,071/day ≈ 21,400 units/month.
So SMED is valuable - but it is not the first move. Raising CNC output while welding is the constraint simply builds WIP in front of a starved-or-saturated welder; the plant would spend three facilitator days and gain zero shipped units.
Correct sequence (constraint first):
- Exploit the weld constraint: attack the 4% rework (each rework consumes a full 61 s of constraint time), restore uptime toward 95% through quick nozzle/torch maintenance, and move all non-welding tasks (fetching, fixturing, minor inspection) off the welder.
- Quantify what is left: rework to 1% and uptime to 95% gives 61 × 1.01 / 0.95 = 64.9 s - still short of 58.7 s, so cycle content must also come down or capacity added (second welder, overtime, offload a weld to a manual cell). Size overtime explicitly: without relief the gap is roughly 2,700 units × 68.9 s ≈ 52 welder-hours per month.
- Then run the SMED event, because once weld is relieved, CNC at 18,000/month becomes the next constraint - which is exactly the argument that justifies the facilitators, made in the right order.
- Subordinate everything else - stop celebrating a never-idle CNC; deliberately idle it rather than overproduce.
The exam point: the kaizen is directionally correct but sequenced wrong, and sequencing is the Black Belt's judgement, not the facilitator's.
3. Where would you place inventory, how much, and what would you measure to confirm the constraint has actually moved?
Read the worked answer
Buffer placement. Put a controlled buffer immediately in front of welding, because welding is the constraint and must never be starved. Do not buffer in front of CNC or after assembly; those buffers only hide problems and lengthen lead time.
Buffer sizing. The buffer must cover the most likely upstream interruption, which is the CNC changeover of 90 minutes. At a constraint demand rate of one unit per 58.7 s, 90 minutes is about 92 units. Set the buffer at 90–100 units with a minimum/maximum marked on the floor, and shrink it as SMED reduces the changeover - this is the visible link between the SMED event and the buffer, and the number to put on the project charter.
After weld relief - check the wash step. A 300-unit batch with a 25-minute cycle creates a step, not a flow: lead time through the wash alone is 25 minutes plus queue, and any defect found after washing is discovered 300 units late. Start by halving the batch, watch the test step (48 s) and assembly (55 s) for the next constraint - both are only just inside the 58.7 s takt and will become visible once weld runs below 55 s.
Measurements that prove the constraint moved:
- Welder OEE (availability × performance × quality) weekly, with rework separated out.
- Buffer level and welder starvation minutes per shift - a buffer that never empties and a welder that never waits means the constraint has shifted; a buffer that empties daily means it is undersized or CNC is still the real problem.
- Throughput in units per day against 920/day, and overtime hours per month - the metric the plant manager actually cares about.
- Do not use CNC utilization as the success measure; a falling CNC utilization with rising throughput is the correct signal here.