Interference Analysis in Manufacturing Cells
Investigate interference between people and equipment using observations, event sequences and capacity data.
Updated:
Simultaneous attendance requests can create queues in a multi-machine cell. Sampling estimates occupancy of these states; it does not automatically determine the optimal machine/operator ratio or minimum cost.
Define causes before attribution
Separate automatic running, manual attendance, confirmed operator waiting, breakdown, missing material, setup and time outside the schedule. A person away from a machine may be addressing another legitimate need. Absence alone does not establish cause.
Record the unit and exposure: machine-time within planned production, equipment group and staffing. Prevent overlap between operator waiting and logistics using clear rules and an unconfirmed-cause code.
Interference share
p̂=operator-waiting readings / eligible readings.
With representative selection and suitable weights, this estimates a time share. It does not directly give episode duration, request sequence or service time; use event records or time study for those questions.
A percentage alone does not establish understaffing. Consider tasks, automatic cycles, demand, travel, safety, breaks and skills. Better sequencing may be an alternative to another person.
Design and precision
Under a simple independent model, n₀=ceiling(z²p₀(1−p₀)/e²). With p₀=0.10, z=1.96 and e=0.03, n₀=385. Three points equal 30% relative to 0.10. If p is unknown, 0.5 gives 1,068 for the same absolute margin.
Machines in a round may share a stoppage. Do not treat every participation as independent information; consider grouping, shifts and periods. Random timing guarantees neither independence nor absence of reactivity.
Simulated example
A cell has three machines and one operator. Sixty of six hundred machine-time readings are attendance waits: 10%. Its interval depends on the design, not just the denominator. Investigate whether waits coincide with changeovers or replenishment.
Safely trial staggered starts, closer tools, separated logistics or assignments. Queueing models or simulation need arrival patterns, attendance times, cycles and variability. The aggregate share does not provide all parameters.
Economics, OEE and energy
Compare incremental cost with contribution margin from genuinely recoverable production. Revenue is not a saving; check demand, bottlenecks and overlapping benefits.
OEE needs planned time, run time, ideal cycle and good/total counts. Waiting shares do not supply every factor. Energy analysis needs state-specific power and representative duration: stopped equipment does not necessarily consume zero or full rated power.
Worksamp scope
The application supports planning, capture and activity distribution. An automatic assignment/minimum-cost optimiser or a “95% accuracy in two weeks” pilot is not established here. Confidence and precision differ.
See Worksamp and complement specific timings with Cronometras when appropriate.