End-of-Line Test Station Design: How to Lay Out a Gas Appliance EOL Station

Start From the Test List

The station should be built from a written test list, not from a convenient layout. For each test — dielectric, insulation, earthing, ignition, flame, FFD timing, internal and external tightness, thermal efficiency, CO, water temperature rise — record the required connection, the instrument, the limit, and the time. Tests that share a connection should be grouped so the appliance is connected only once, and tests that depend on a prior result should be ordered so the station cannot skip the safe step. A test that needs a live flame should not be reachable before the dry and electrical checks, and a gas test should not be reachable unless the exhaust is proven.

Connections the Station Must Provide

Sequence and Safety

1. Connect the appliance, scan the serial and confirm the station is clear and exhausted;
2. Run the dry electrical checks (dielectric, insulation, earthing);
3. Admit gas, prove the exhaust and run ignition and flame;
4. Measure FFD timing, tightness and combustion;
5. Run the water test for water heaters and boilers;
6. Record the result and release only on a full pass.

The station should be physically safe: the operator should not reach across a live flame or gas connection, and a failed unit should be held in a defined position rather than returned to the pass conveyor. Gas isolation should be positive and automatic, and the station should fail safe on lost air, power or exhaust. Interlocks are described in G7 and calibration in G9.

Fit the Station to the Takt

The number of stations follows the takt and the cycle time, not the number of tests. If a full test cycle is longer than the takt, a single station becomes the bottleneck and the line backs up; the correct response is to split the sequence over stations or provide parallel stations, so each station’s work fits within the takt. This is covered in G3, with inline versus offline choices in G4. A station that is rushed to fit the takt will either skip tests or produce inconsistent readings, which is the common reason a station appears to work but releases defects.

Common Design Mistakes and How They Appear in Production

Most EOL stations that fail in service do not fail because a single component was wrong; they fail because a connection was assumed rather than built. A station that shares the workshop gas supply without a local regulator, for example, will show combustion readings that shift whenever another line turns on, and the operator will see intermittent failures that no one can explain from the appliance itself. A station that draws its electrical test from the same circuit as the rest of the workshop will occasionally trip or read a lower voltage during a heavy start elsewhere, and the unit will be blamed for a condition that belonged to the supply. These are the patterns that look like random appliance defects for weeks until the station is finally measured against a fixed reference. The correction is to give each station its own regulated pressure, a protected electrical circuit, filtered water, and a proved exhaust, so the condition of the test is independent of the surrounding factory.

A second group of mistakes comes from treating the station as a fixture rather than as a gate. A station that allows the operator to start without a serial, to release a unit while the record is still open, or to override a failed test with a reset, has effectively no interlock even though it displays the correct readings. The failures it produces are not caught until the appliance is in the field, and the factory cannot then show which unit was tested. A third group is the opposite fault: a station that is made so elaborate that its cycle exceeds the takt, at which point the line begins to rush or to skip tests to keep pace. The practical checks are the same in every case — the station must hold its own stable conditions, must physically gate the release, and must complete the full sequence within the takt. If any of these is missing, the station is a measurement fixture rather than a release gate, and the missing function should be corrected before the line is run rather than worked around by the operators.

A Final Acceptance Checklist

• Each connection (gas, water, electrical, exhaust, data) has its own controlled condition and a proof where safety is involved;
• The sequence is fixed, starts only after a valid serial, and runs the full list in a safe order;
• The cycle, including slow units, is measured and at or below the takt;
• Each failed condition holds the unit, closes the gas where required, and routes to a repair path;
• A new operator can run the station after a short instruction and a repair unit returns to a retest.

How the Five Connections Are Commissioned and Proven

The connections are only as trustworthy as the way they are first proven, so commissioning should be carried out connection by connection rather than by running the whole sequence once and accepting it. The gas connection is proven first by setting the local regulator to the required pressure with a reference gauge and then by deliberately changing the workshop supply to confirm the station holds its own pressure; the water connection is proven by flowing through the filtered, regulated path and checking that the measured flow remains stable under the conditions the line actually uses; the electrical connection is proven at the destination voltage and frequency on a protected circuit, with the safety tester verified against its own reference; the exhaust connection is proven as a gate, because a live flame should not be permitted unless the extraction is actually moving air; and the data connection is proven by checking that the serial opens the correct sequence and that the result reaches the record even under a simulated network loss. Each connection is deliberately faulted in turn and the station must move to its safe or held state rather than continuing. A station that passes these five fault tests is a genuine release gate; one that has only run a clean sequence on good units is unproven, because the first marginal supply or failed sensor in production will then be its first real test.