Generator Stator Air Tightness Test

The Role of Leak Testing in Hydrogen-Cooled Generators

In large turbo generators, air tightness testing is not just a simple "did the pressure drop?" check. Especially in hydrogen-cooled machines, this test is the practical way to see whether the generator can safely contain the gas. In high-power generators used in gas turbine, steam turbine, and combined cycle power plants, significant heat is generated in the stator and rotor areas. This heat must be removed steadily; otherwise, insulation ages, winding temperatures rise, and load capability decreases.

Hydrogen is not chosen for nothing. Its thermal conductivity is significantly higher than that of air. Roughly speaking, hydrogen's thermal conductivity is around 0.168 W/(m·K), while air is about 0.026 W/(m·K). That means hydrogen transfers heat much more easily than the same volume of air. Moreover, its density is very low. At 0 °C and 1 atm, hydrogen density is about 0.09 kg/m³, while air density is about 1.29 kg/m³. When the rotor rotates, friction and windage losses occur in the gas; low density provides an advantage here.

In the field, this means: hydrogen both cools well and reduces gas resistance for rotating parts. In large machines, this difference is felt in efficiency, temperature, and loadability.

But hydrogen has a tough side too. When mixed with air in certain proportions, it can form a flammable and explosive mixture. The presence of hydrogen in air between about 4% and 75% is considered risky. Therefore, before hydrogen filling, the casing, connections, seals, cooler area, seal oil system, and gas circuit are tested as reliably as possible. It's not just about the gas bill; it's about safety.

Why is the test often done with air?

After commissioning or maintenance, the first check is usually not done with hydrogen. The system is first pressurized with air or, depending on the procedure, a suitable inert gas. The aim is to see whether the casing and gas circuit hold pressure. If the test is acceptable, purge steps are applied; air is removed, the necessary intermediate gas is used, and the system is prepared for hydrogen operation.

This sequence may vary by manufacturer. In documents from manufacturers like GE, Siemens, Alstom, or Mitsubishi, volume, test pressure, holding time, acceptance limit, and purge steps may be defined differently. Still, the practical logic remains: before hydrogen is introduced, the system's gas retention behavior is observed.

If leakage is high while hydrogen is present, several things worsen at once. Gas consumption increases. Hydrogen purity may drop. The seal oil system is stressed more. In some areas, there is a risk of hydrogen accumulation. If moisture and air ingress occur, cooling performance also degrades. Then the question "why is the generator running hot?" arises. Sometimes the answer is not on the electrical side but on the gas side.

The Gas Logic Behind the Measurement

In leak tests, pressure difference seems like the most straightforward data at first glance. If you see 300 kPa at the start and 290 kPa at the end, you say "it dropped 10 kPa." But this statement alone is incomplete. Because gas pressure changes with temperature.

If the gas cools, pressure drops. If the gas heats up, pressure rises. Even if there is no leak in the system, a temperature change creates movement on the manometer. Therefore, in stator leak calculation, pressure must be read together with temperature, not in its bare form.

The basis of the calculation is this ratio:

Here, stator pressure is not used alone; it is evaluated together with atmospheric pressure. Because the stator pressure read in the field is usually gauge pressure. That is, the value above atmospheric pressure. To approach absolute pressure, atmospheric pressure is added:

In a plant near sea level, the approach of "just use 100 kPa for atmospheric pressure" sometimes doesn't cause much trouble. But at a high-altitude site or in a sensitive acceptance test, this habit spoils the calculation. If the barometric pressure is actually around 88-90 kPa, entering 100 kPa is not an innocent rounding.

The same attention is needed on the temperature side. If the stator casing is still cooling after maintenance, a test performed before temperature equilibrium is reached measures thermal transition more than leakage. On paper, the formula works correctly; if the data is bad, the result is still bad.

Can leakage occur even when pressure rises?

Yes, in some examples, the final pressure appears higher than the initial, but because the temperature effect is greater, the calculation may still show gas loss. The opposite is also possible. Pressure drops a little, but temperature also drops; in this case, the actual leak may be lower than the bare pressure difference.

Therefore, making a decision by only writing pressure on the test form is hasty. Pressure, atmospheric pressure, temperature, and time must be read together. That's the logic of the test.

Data Used in the Calculation

The stator gas volume represents the enclosed volume under test. In the calculator, it is noted that with the rotor in place, 110 m³ is typically used, and without the rotor, 120 m³ can be used as a help hint. This is a good practical reminder, but it is not an exact value. Even in two generators of the same power class, casing volume, cooler arrangement, and internal geometry can differ. The volume stated on the manufacturer's test sheet should be used.

The time interval also directly affects the result. Short tests amplify measurement error. For example, in a 10-minute test, a 1 kPa reading difference can turn into a large daily leak value. A longer, more stable test of several hours generally appears more reliable. Of course, here too, the system temperature must have stabilized.

For start and end, three values are taken: stator gauge pressure, atmospheric pressure, and stator temperature. These must be read at the same time or very close to each other. In the field, sometimes pressures are written first, then temperatures are taken by walking around. If the machine temperature is changing, even this small delay affects the test record.

Care must be taken with pressure units. If absolute pressure is entered in the stator pressure field, atmospheric pressure is added a second time. This error inflates the result. The calculator expects stator pressure as gauge pressure; atmospheric pressure is entered separately.

For temperature, representativeness is also important. In large casings, a single-point temperature may sometimes be insufficient. If the test procedure requires multiple temperature points, an average or defined representative temperature should be used. The convenience of "let's just put 20 °C" is not always correct.

Formula and the 110 m³ Scenario

The main formula in the calculator is structured as follows:

In this formula, \(Q\) gives the gas leak rate. \(V_1\) is the stator gas volume. \(P_1\) and \(P_2\) are the stator gauge pressures at the beginning and end of the test. \(B_1\) and \(B_2\) are the atmospheric pressures at the same moments. \(T_1\) and \(T_2\) are the start and end temperatures. \(h\) is the test duration.

Consider a generator with a volume of 110 m³, assumed to have the rotor installed, and we perform a 1-hour test. Let the initial stator pressure be 300 kPa, atmospheric pressure 100 kPa, and stator temperature 20 °C. At the end, let stator pressure be 290 kPa, atmospheric pressure still 100 kPa, and temperature still 20 °C. The first check is:

Since temperature remained the same, this example is relatively easy. The pressure drop is directly interpreted as gas loss. However, in the real field, most tests are not this clean. The start might be 23 °C and the end 27 °C. Or the casing might slowly cool during the test. That's why the calculator includes temperature in the formula.

In the given test scenario, the stator volume is 110 m³, duration 7.5 hours, initial values P₁ = 502.61 kPa, B₁ = 100.38 kPa, T₁ = 23.75 °C; final values P₂ = 507.6 kPa, B₂ = 100.3 kPa, T₂ = 26.72 °C. Notice that the final stator pressure is higher. Yet, because the temperature also rose, the gas quantity interpretation cannot be made by looking only at the pressure increase. This is a good test example; it keeps one from relying on rote.

Field Interpretation of the Result

When the gas leak rate is converted to a value like m³/day, it becomes easier to compare tests performed over different durations. It would not be correct to place a 2-hour test and an 8-hour test side by side using bare pressure difference. The daily leak equivalent is useful for this reason.

Still, the result should not be expected to make the decision alone. The acceptance limit is defined in the manufacturer's procedure, plant operating instructions, or maintenance test plan. A leak value considered acceptable in one generator may be considered excessive in another machine. Casing volume, hydrogen operating pressure, seal oil arrangement, cooler configuration, and machine age affect the result.

In hydrogen operation, if leakage is high, gas consumption increases first. The operator notices this from the make-up hydrogen demand. But the real issue is safety and purity. When hydrogen purity drops, cooling performance degrades. If moisture is present, an undesirable environment for insulation forms. If there is a problem on the seal oil side, hydrogen may mix into the oil system, or air/gas crossovers may occur in the reverse direction.

Leak points are often not searched in a single location. Terminal boxes, cooler connections, flange surfaces, manhole covers, seal oil areas, gas dryer connections, and measurement instrument lines are checked. Experience shows: sometimes what you think is a big leak comes from a small instrument fitting. Sometimes all the small leaks add up and disturb system behavior.

Most commonly missed details in testing

• Barometric pressure is underestimated. Especially in high-altitude plants, this habit distorts the result.
• The test is started before temperature stabilizes. If the casing is still heating or cooling, the calculation deviates directly.
• Stator volume is entered as an assumption. The difference between 110 m³ and 120 m³ is about 9%; the result changes by the same amount.
• Absolute pressure is entered instead of gauge pressure. This error is silent but spoils the result.
• Start and end values are read with different discipline. The reliability of a test comes from the measurement arrangement before the formula.

Brief Conclusion

Generator stator leak testing, especially in large hydrogen-cooled generators, is a safety check step before hydrogen filling. The purpose of the test is not only to see the pressure drop but to calculate the gas leak rate by reading the effects of volume, duration, atmospheric pressure, and temperature together.

Hydrogen is a good coolant, reduces losses around the rotor, and provides significant advantages in large machines. But if leak control is weak, this advantage quickly turns into safety and operational problems. Therefore, the test result should be evaluated together with measurement quality and manufacturer acceptance limits. The formula is a tool; solid data provides the decision.