Assembly Line Balancing (RPW)

How Ranked Positional Weight Line Balancing Works

Assembly line balancing distributes work elements across workstations so every station fits within the target cycle time (takt) while respecting precedence — a task cannot start before its predecessors. The goal is the fewest stations with the most even load.

The Ranked Positional Weight method (Helgeson & Birnie, 1961) scores each task with its own time plus the times of all tasks that must follow it. Tasks are then assigned in descending weight order to the earliest station that has both capacity and all predecessors already placed — tasks may share a station with their predecessors, because the within-station work sequence preserves precedence.

Three measures summarize the result: line efficiency = total task time ÷ (stations × cycle time); balance delay = 100% − efficiency, the share of paid time spent idle; and the smoothness index, the root-sum-square deviation of station loads from the busiest station — 0 means a perfectly even line.

When to Use This Tool

• Designing a new assembly line when you know task times and precedence from a process chart
• Rebalancing an existing line after demand (takt time) changes, before physically moving equipment
• Estimating headcount: stations × operators per station for a target output rate
• Checking whether a proposed cycle time is feasible — any single task longer than the cycle time is an immediate red flag
• Comparing balance quality between two candidate cycle times using efficiency and smoothness index

Common Mistakes to Avoid

• Using machine cycle time instead of operator task time — balance the constraint resource you actually staff
• Omitting precedence links because "everyone knows the order" — missing links let the heuristic create assignments the floor cannot run
• Setting cycle time equal to takt time with no allowance — real lines need 5-15% margin for variation, fatigue, and micro-stops
• Splitting a task across two stations on paper when it is physically indivisible — split tasks in the data only if the work itself can be split
• Reading a heuristic result as the provable minimum — RPW is fast and good, but exact methods can occasionally save a station on hard precedence networks

How to Interpret Results

• Stations vs Theoretical Min: matching the lower bound ⌈Σ time ÷ cycle time⌉ means no better station count exists; one above is typical for constrained networks
• Line efficiency above 85% is generally considered a well-balanced line; below 75% indicates fragmented tasks or a poorly chosen cycle time
• Balance delay is the idle share you pay for — 12% balance delay on a 10-operator line is more than one full operator of idle time
• Smoothness index compares evenness between alternatives with the same station count — lower is better, 0 is perfectly level
• A station loaded exactly to the cycle time has zero buffer; consider shifting a small task off the bottleneck station for robustness

Related Standards & References

• Helgeson & Birnie (1961) — Assembly line balancing using the ranked positional weight technique (Journal of Industrial Engineering 12:394-398)
• Salveson (1955) — The assembly line balancing problem (Journal of Industrial Engineering 6:18-25), the first formal statement of SALBP
• Scholl & Becker (2006) — survey of exact and heuristic SALBP methods (European Journal of Operational Research 168:666-693)
• MTM / MOST predetermined time systems are the usual source of task times fed into line balancing

Frequently Asked Questions

What cycle time should I enter?

Start from takt time (available time ÷ demand) and subtract an allowance for variation and minor stops, typically 5-15%. The cycle time must be at least as long as your longest indivisible task, otherwise the problem is infeasible.

Why does the station count exceed the theoretical minimum?

Precedence constraints and indivisible task times often make the lower bound unreachable — e.g. two 6-minute tasks cannot share a 10-minute station even if total time suggests they could. One station above the bound is common and usually acceptable.

How do I enter parallel branches?

List every task once with its own predecessors; branches simply share a predecessor and rejoin where a later task lists both branch ends. The RPW weights automatically prioritize the longer branch.