StudyCorner
65 / 78

Lesson 65 of 78 · Factory Flow & Data

Throughput, Bottlenecks, Buffers & OEE

MIT’s manufacturing-systems course analyzes material and information flow, capacity, queues, bottlenecks, buffers, batches, schedules, and dynamic behavior 1.

Three production stations with different effective rates are separated by buffers, and a timeline shows blocking, starvation, and downtime
System throughput emerges from interacting rates, variability, failures, setups, quality, and buffer policy—not the fastest machine's cycle time. Credit: StudyCorner original diagram · CC BY 4.0 · Source

Capacity is not nameplate speed

A nominal 12-second station can make 300 units/hour only with no setup, failure, starvation, blocking, or scrap. If availability is 90%, performance while running is 95%, and first-pass quality 98%, the familiar OEE product is

\[ 0.90(0.95)(0.98)=83.8\%. \]

Equivalent good output would be about 251 units/hour. Use OEE components as loss categories, not a contest that motivates hidden downtime or reclassified scrap.

Little’s Law

For a stable flow over a consistent boundary,

\[ WIP=Throughput\times Flow\ Time. \]

At 40 good units/hour and average 15 units in the system, average flow time is 0.375 h or 22.5 minutes. Automation that raises work-in-process without raising throughput lengthens lead time.

Blocking, starvation, and buffers

An upstream station is blocked when it cannot release output; downstream is starved when it lacks input. Buffers decouple short variations but consume floor, working capital, traceability effort, and response time. Size them from failure/repair distributions and policy, not a round number.

Improve the constraint

Measure constraint uptime, cycle distribution, setup, microstops, quality, and feeding. Protect it from starvation and blocking; move nonessential work away; reduce its variability; then reassess because the constraint can move.

Flow model

Build a spreadsheet or discrete-event simulation for three stations with random downtime and a finite buffer. Compare zero, small, and large buffers. Report good throughput, WIP, blocking, starvation, and lead time. Recommend a buffer only after showing what loss it addresses.

Source trail

References

  1. 1
    Introduction to Manufacturing Systems. MIT OpenCourseWare. verifiedMaterial and information flow, capacity, queues, buffers, batches, bottlenecks, scheduling, optimization, and production-system dynamics. Cited at: course scope.
Further reading
  • Design and Manufacturing II. MIT OpenCourseWare. verifiedModern manufacturing organized around process physics, equipment and control, manufacturing systems, and design for manufacture.

Check your understanding

  1. What normally limits the long-run throughput of a serial line?
  2. What can a buffer do?