1.5 Assembly Flow Design

Assembly flow design is where manufacturing turns from theory into rhythm. By matching takt time to real, balanced work and shaping the layout for both speed and ergonomics, a line can run steadily instead of lurching from fire to fire. Whether built as a U-cell for flexibility or a conveyor for volume, flow only stabilizes when tests act as the pacemaker and bottlenecks stay fed and visible. The result is not just higher output, but calmer operators and fewer surprises.

1.5.1 Goal (in one line)

Build a line that moves at customer pace (takt) with no drama: parts arrive when needed, hands travel little, and the slowest step is fed, free, and visible.

1.5.2 Start with the math (tiny, honest, and powerful)

Available time/shift = shift minutes − breaks/meetings/cleanups.
Takt time = Available time ÷ Required units.
Cycle time (per station) = real average seconds to do the work, including normal micro-stops.

Design so every station’s cycle ≤ takt, with the bottleneck just under takt and protected.

Example

Available: 7.5 h = 27,000 s. Demand: 300 units → takt = 90 s/unit.

Split work so each station ≈ 80–88 s and the bottleneck has a small buffer ahead.

1.5.3 Pick a flow pattern (use the lightest one that works)

Pattern	When to use	Pros	Watch-outs
Single bench	L1–L2 products, low volume	Minimal WIP, flexible	One person = one rate; skill variance
U-cell (1–6 ops)	High-mix, medium volume	Short walks, shared tools, easy help	Needs cross-training; bottleneck can starve without a small buffer
Inline (serial)	Medium–high volume, clear sequence	Simple pacing, easy to see flow	Long walks if poorly arranged; changeovers hurt more
Conveyor / pulse line	High volume, repeatable	Fixed pitch, easy station timing	Less flexible; rework loops needed
Hybrid (U feeding conveyor)	Mixed families + runner SKUs	U does prep; conveyor finishes	Two rhythms to manage

Rule: If you change models often, start with a U-cell; if you ship one product all day, build an inline/pulse.

1.5.4 U-cell basics (make it flow in a small footprint)

Arrange stations in a ∪ so the product and eyes travel clockwise, materials inside the U, finished goods out the open end.
Two-person U: split work 60/40; the faster operator floats to help the bottleneck.
Pitch board at the exit shows planned vs actual every 30–60 min.
Keep shared tools (label printer, torque driver presets) at the base of the U.

Feeding the U: module supermarket (22.3) behind operators; carts roll into the U on casters, one cart = one WO/Variant.

1.5.5 Design the line in eight moves (whiteboard to floor)

List work elements with honest times (stopwatch 5–10 units).
Group by skills & tools (torque set, adhesive, programming).
Balance with a Yamazumi (stacked bar) until each station ≈ takt.
Pick layout (bench/U/inline). Sketch reach zones and walking paths.
Place materials: inside the U or right side of inline; heavy parts at waist, fasteners in color cups by torque group.
Add micro-buffers: WIP shelves before/after the bottleneck (1–2 units).
Gate tests: put the pacemaker (functional/safety test) near the end; everything flows to it.
Run a pilot: 10–20 units, time each station, fix the tallest Yamazumi bars, repeat.

1.5.6 Balancing tactics (fast wins)

Split the pile: move one or two heavy elements from the slow station to neighbors.
Parallelize: build PSU tray/fan wall/display door as L1 modules (22.3) off-line.
Change the unit: build in pairs (two units per pitch) if fixtures/tool change time dominates.
Kit smarter: move screw hunting into kitting; arrive in torque groups.
SMED at changeover (18.3): pre-load labels, images, torque maps by SKU scan; swap fixtures on zero-point pins.

Mini example (balancing by minutes)

Four stations, takt 90 s. Times: 120 / 70 / 75 / 60 → Station 1 is the bottleneck.

Move “fan wall install” (30 s) to Station 3 → new times: 90 / 70 / 105 / 60.

Then move “label set” (20 s) from 3 to 4 → 90 / 70 / 85 / 80. Done.

1.5.7 Buffers & pacing (Little’s Law without the lecture)

Size buffers by variation: start with 1–2 units before/after the bottleneck.
Use FIFO lanes with visible max lines; overflow means you just found a problem.
Pitch (fixed release every takt) keeps rhythm: a small timer beeps; if red lights persist at one station, rebalance.

1.5.8 Ergonomics & reach (speed = comfort)

Hands work in the elbows-down zone; heavy picks between knee and chest.
Two-handed tasks? Put bins split left/right to avoid crossing.
Torque drivers on retractors; bits parked in a labeled shadow.
Lighting: bright, diffuse; avoid glare on glossy plastics and label windows.
Turntables / tilt stands to rotate the chassis instead of the operator.

1.5.9 Material presentation (zero hunting)

One cart = one unit (or wave); variant color band on handle.
Fasteners by torque group (color cups) right at the station.
Consumables kits: TIM syringe with bead size card; threadlocker dots by color.
Chokepoints (programming, label print) centralized with queue visibility.

1.5.10 Tests, rework loops, and the pacemaker

Put the functional/safety test near the end; it becomes the pacemaker.
Rework loop off-line: failed units exit to NG-QUAR/REWORK without blocking flow.
If test time > takt, add two testers in parallel or run burn-in as a side loop.

1.5.11 Changeover planning (keep flow during mix)

Heijunka (leveling): fix a daily product wheel so shared setups stick around.
Changeovers are internal only for the bottleneck; convert all else to external prep (carts, labels, images).
Measure CO time = last good → first good; target ≤ 5–10 min for mature cells.

1.5.12 Control & visibility (what the cell board shows)

Takt vs actual count, updated each pitch.
Starved/blocked minutes per station (pinpoint the constraint).
Top 3 stop reasons with one countermeasure in progress.
Quality at source: misses by station (labels, torque, routing) with a tiny Pareto.

1.5.13 Metrics that prove the layout works

Throughput = units/shift at or above plan.
OEE (18.4/16.3) with focus on Performance leg (cycle time creep).
Changeover time and first-good after change count.
Walk distance per unit (aim < 10–15 m in a U-cell).
WIP before bottleneck (should hover near target, not explode).
Defects at source trending down (labels, torque, routing).

1.5.14 Common traps → smallest reliable fix

Trap	Symptom	First move
Layout by gut, not takt	Nice pictures, poor output	Do the takt math, build the Yamazumi, then draw
Bottleneck starved	Idle key station	Add 1–2 unit buffer before it; rebalance elements
Long walks for parts	Tired operators, slow pace	Bring materials inside the U; shadow tools; split bins
Shared tools in the wrong place	Queueing for torque driver	Duplicate or move tool to the bottleneck; preset bits
Test is the unseen constraint	Units pile at test	Parallel testers; make test the pacemaker; reduce dwell
Changeover thrash	Output dips each SKU swap	Pre-stage carts; zero-point fixtures; MES push of recipes
Rework blocks the lane	Line stops on failures	Create a side rework loop with NG-QUAR state

1.5.15 Pocket checklists

Design (whiteboard)

Demand → takt calculated
Work elements timed; Yamazumi balanced near takt
Flow pattern chosen (bench/U/inline); sketch reach & walks
Bottleneck identified; buffers sized (1–2 units)

Pilot (first 20 units)

Station cycles measured; tallest bar reduced
Materials reachable; torque/label tools placed well
Test station sets the pace; rework off-line

Daily run

Pitch board green; starved/blocked < 15% at bottleneck
Changeovers ≤ target; first-good in ≤ 2 units
One small kaizen moved from board to standard each shift

Conclusion: Designing lines with takt-driven math, balanced workloads, and clear pacing transforms assembly from reactive to predictable. With smart kitting, ergonomic layouts, and disciplined buffers, production achieves flow that is both efficient and sustainable.