Production flow & line balancing

At this event state: 3 orders in the system; 6 flow-complete; 0 modeled scrap exits. Totals include warm-up; headline metrics exclude it.

Candidate minus reference output: 0.00 skids / operating week; paired 95% CI 0.0–0.0.

Reference uses one assembly worker, one kit worker, one test pair, two waiting slots, four-valve maximum batches, and baseline layout/handling. Both policies use the same demand, mix, release pattern, variation, scrap probability, transfer speed and observation window. Paired differences use matching job-level random inputs. Intervals estimate simulation replication means; they are not real production guarantees or prediction intervals.

Work and blocking are integrated over event intervals after warm-up, divided by resource capacity × measurement minutes. A blocked resource cannot start another job. T1 working fraction measures occupied test stands, not technician effort; each stand reserves a dedicated technician during the cycle. Highest selected-run working fraction: A1. This is a diagnostic clue, not proof of a bottleneck under every policy.

Standard routing · released at 0.0 operating min · Production-complete flow exit · total ready-queue wait 60.0 min. Flow lead time 516.3 min.

Every kit consumes one valve, one harness set, two pumps, one heat exchanger and one frame. Selected-run kits started: 40; purchased consumption: 40 harness sets, 80 pumps, 40 heat exchangers and 40 frames. Raw valve allocations: 40 = 40 machined + 0 not yet machined.

Purchased material for these newly generated orders is assumed available and independently dispositioned. No original SYN-B1/B2 supplier stock or SPC observations are used as inventory. Scrap is a fictional terminal skid failure rule with no rework loop; it is independent of ML scores and existing incident/FAT evidence. Completed-cohort lead times can understate waiting in overloaded scenarios because unfinished orders are excluded.

Problem & scope

Investigate which staffing, task assignment, batch and buffer policies support a declared demand under hypothetical process variability.

Responsibility & approach

This independent demonstration combines a precedence-aware line-balancing search, event simulation, route geometry, work assumptions and statistical comparisons. It does not represent an employer production deployment.

Decisions & implementation

Separate valve batches from skid units; reserve finite waiting space before transport; retain blocked output on its upstream resource; simulate complete orders; exclude a declared warm-up and inspect every flow transition.

Results & lessons

The numbers above are generated by the active scenario. More assembly workers can move the limiting resource to test; tighter buffers can block otherwise available resources. Compare paired intervals, inspect queues and account for unfinished work before treating a policy as beneficial.

Selected route basis: baseline; loaded route lengths M1→K1 17 m, K1→A1 5 m, A1→T1 11 m, T1→D1 2 m. Transfer time = route metres / (60 × 1 m/s). Machined outputs may split into smaller bin transfers when waiting space is scarce. Assembly-cell transfers inside A1 have zero modeled distance. Handling element: 18 s / equal 10 kg payload.

Dedicated unconstrained transport capacity is assumed; empty returns, congestion, breaks, shifts, breakdowns, receiving/quality resource queues, supply shortages, repair/rework and calendar deadlines are omitted. Warm-up reduces empty-start bias without proving steady state. An overloaded queue can continue growing. Workstation timing is a work-study assumption; a lifting index is never converted into speed or a safety approval.