Line balancing and Yamazumi chart

How line balancing works

Line balancing shares the tasks of an assembly line between stations so that every station finishes within takt and the stations are loaded as evenly as possible. Each idle second at a station is paid for on every unit; each second over takt is lost output. The Yamazumi chart stacks the tasks of each station so both show at once.

Measures

• Work content Σt = the sum of all task times. Theoretical minimum stations = ⌈Σt / takt⌉.
• Line efficiency = Σt / (k × S_max), with k stations and S_max the longest station time (the line's cycle time). Balance delay = 1 − line efficiency: the share of paid station time that is idle.
• Smoothness index = √Σ(S_max − S_j)²: 0 when all stations take the same time.
• Efficiency at takt = Σt / (k × takt): how well the stations use the time the customer pace allows.
• Precedence: a task's predecessors must be done at the same station or an earlier one. Stations are ordered as they first appear in the data; a task placed before one of its predecessors is reported.

The proposal

• Ranked positional weight (Helgeson and Birnie, 1961): each task's weight is its own time plus the times of every task that follows it, directly or through others. Tasks are ranked by weight (ties: the longer task, then the order entered).
• Largest candidate rule: tasks are ranked by their own time (ties in the order entered).
• Stations are then filled one at a time: going down the ranking, the first task whose predecessors are all assigned and which still fits within takt joins the station; when no task fits, the next station opens.
• Both are single-model heuristics: they need every task to be shorter than takt, and do not guarantee the fewest stations. “Use the proposal as the station column” writes it into the data, where it can be moved by hand.

Worked example (the tool's example)

1. Twelve tasks of a coffee grinder assembly take Σt = 252 s. With a takt of 60 s at least ⌈252 / 60⌉ = ⌈4.2⌉ = 5 stations are needed.
2. As entered, the five stations take 44, 56, 62, 58 and 32 s. Station S3 (wire motor, hopper, connect: 62 s) is over takt. Line efficiency = 252 / (5 × 62) = 81.3 %, balance delay 18.7 %, smoothness index √(18² + 6² + 0² + 4² + 30²) = 35.7.
3. The ranked positional weight proposal also uses 5 stations: 50, 50, 50, 58 and 44 s (base and motor; wire motor and burrs; switch, connect and hopper; housing and test; label, pack and fetch parts). The longest is 58 s, within takt: line efficiency = 252 / (5 × 58) = 86.9 %.
4. The chart also shows that “Fetch parts” (12 s) is waste and test, label and pack (62 s) are necessary but add no value: moving parts closer removes a block instead of moving it.

Common mistakes

• Balancing to takt with no margin: every disturbance then stops the line. Balance to a planned cycle time below takt.
• Using standard times that were never observed; time the tasks on the line, several cycles each.
• Forgetting walking, fetching and waiting: they belong on the chart as waste, so that they can be removed.
• Ignoring precedence or zoning (one side of the product, a fixed tool, a test booth) that the heuristic cannot see.
• Judging a balance only by the number of stations: two lines with 5 stations can differ a lot in their longest station.

Input and files

• Columns are found by their header names: Task, Time, Station, Predecessors, Category. Predecessors are task names separated by commas or semicolons (quote the cell in a comma-separated file). Category: VA, NVA (or NNVA) or Waste; anything else is drawn plain.
• Files: CSV, TSV and TXT are read as if pasted (UTF-8, UTF-16, otherwise Windows-1257). XLSX: choose the sheet and the columns or a cell range; formula cells use the value Excel last saved, dates are kept as text. Files up to 20 MB are read in this browser only and never uploaded.
• XLSX files are read with SheetJS Community Edition 0.20.3, Apache License 2.0 (license text at the end of this file).

Questions

What is a Yamazumi chart?

A stacked bar chart with one column per station (or operator) in which each block is one task, drawn to its time. Yamazumi is Japanese for “piled up”. With the takt line across it, it shows at a glance which stations are overloaded, which have idle time and how much of the work adds value.

What is a good line efficiency?

It depends on how finely the work can be split; 85 to 95 % is common for a manual assembly line. Efficiency is measured against the longest station, so a line can be well balanced and still too slow: check the longest station against takt as well.

Does the ranked positional weight method find the best balance?

Not always. It is a heuristic: it usually gets close to the minimum number of stations but does not guarantee it, and it ignores zoning, reach, tools and operator skills. Treat its result as a proposal to check and adjust on the line.

What if one task is longer than takt?

No assignment of whole tasks can meet takt. Split the task, run two stations in parallel for it, speed it up, or increase the available time. The proposal is not made until every task fits takt.