How Many Test Stations Per Production Line: Matching Stations to Output

The Calculation

Parallel vs Sequential Stations

There are two different ways to add stations. Parallel stations each run the same full test, which is the right answer when the test cycle exceeds the takt but the sequence is indivisible; sequential stations split the test, so the first station performs electrical checks while the second performs gas and combustion, which is the right answer when the sequence can be divided and each part fits the takt. The choice depends on whether the bottleneck is the whole cycle or a group of tests, and on how the appliance moves between stations. Splitting tests often needs less duplicated equipment than running several full stations, but it only works if each split station stays within the takt.

Allow for Repair, Re-test and Availability

• A failed unit must leave the production station and follow a separate repair path, so it does not block the line;
• Re-test capacity is required after repair, otherwise repaired units queue behind production;
• Station availability is below 100% — tooling changes, minor stoppages and maintenance reduce the real capacity, so the count uses an availability allowance rather than the ideal time;
• First and last units and changeover need time, which should be counted for short runs.

Lines that are counted on ideal times and zero failures are the lines that later skip tests under pressure. The method links to G1, G3, and the rotary/inline choice in G10, with release control in G7.

Worked Example: A Single Line Taken Through the Count

A factory runs one shift of 8.5 hours and allows 1 hour for breaks, meetings and a planned maintenance window, leaving 7.5 hours, or 27,000 seconds, of available time. It must ship 600 good units per shift, so the takt is 27,000 divided by 600, which is 45 seconds. The test sequence on the appliance is timed over several days and includes slow units; the dry electrical checks take 25 seconds, the exhaust and ignition proofs take 30 seconds, and the live-flame and tightness section together take 85 seconds. The first two groups can run at one position, but 85 seconds cannot fit inside 45 seconds, so the flame and tightness section either needs two parallel stations or has to be split. Because the section is a single live sequence that cannot be divided into two safe halves, two parallel stations are used; the two positions together absorb the 85-second cycle, and each unit still gets one index every 45 seconds.

The count is not finished at this point. A small proportion of units fail and must leave the station, be repaired and return to a retest, so the line is given an availability allowance and a separate repair position rather than counting those units on the production stations. If the first weeks show the live-flame section drifting toward 95 seconds on marginal units, the assumption that produced two stations is wrong and a third parallel position or a small reduction in output is required; the line must not meet the drift by rushing the test. This example shows the order in which the calculation is actually used: takt sets the pace, measured cycle sets the stations, and the repair and allowance set the real capacity. A station count built this way will release close to the target and keep the full test; a count built from the fastest ideal time will appear to work until the first marginal unit, then begin skipping tests or building a queue.

How the Count Changes for Two or Three Shifts and for Growth

The station count is tied to a stated shift pattern, so the same line serving a different number of shifts changes the answer rather than the number of stations changing automatically. A factory that needs 600 units per shift across two shifts has the same takt and the same station count as a single-shift factory that needs 600, because the required output and available time are divided per shift; but a factory that must cover 1,200 units in one shift has half the takt and needs roughly twice the test capacity. The count should therefore always be written against a clear statement of shifts, available time per shift, and good-unit output rather than a daily total that hides the shift pattern. A second practical question is whether the line is expected to grow; a station that is built to the exact count with no positions left cannot absorb a takt reduction, while a line that reserves a position can grow without a rebuild. Growth should be planned deliberately: a line that may move from 600 to 800 units should be given a layout that can add a parallel station or split a step rather than being rebuilt, but it should not be overbuilt with unused positions that idle for the first year. The practical approach is to fix the count for the current output, reserve the mechanical and control space for the next step, and leave the full installation until the output actually moves. This keeps the first investment honest while avoiding a replacement later. The same rule applies to the repair path: as output grows, the re-test capacity must grow with it rather than being treated as a fixed overhead, otherwise a higher failure rate will consume the production stations and quietly reduce the real output below the planned number.

Finally, the count should be treated as a design parameter that is revisited rather than a one-time answer. The first weeks of a new line usually reveal where the real cycle differs from the estimate and where the failures concentrate, and those observations should feed back into the count: a station that is repeatedly above the takt on slow units is telling the factory that the capacity is short, and the correction should be made deliberately rather than absorbed by rushing. A line that is recalculated from measured data in this way converges quickly on a stable count and keeps the full test, while a line that keeps the original number after the conditions have changed is the one that later begins skipping tests.