Detection Of Cooling-Water Intrusion Into Standby-Power Diesel Engines

This case study discusses pitfalls associated with the condition-monitoring of oil in a generator’s lubrication system.

By Randall Noon, P.E.

Diesel-engine generators, the stalwart mainstays of standby-power systems, offer several advantages: They efficiently provide electrical power, when needed, with the push of a button or automatically, perhaps, when under-voltage conditions in the grid trip their start relays. Also, the handling and storage of diesel fuel presents fewer safety concerns than gasoline or natural gas. And finally, an on-site storage tank full of diesel fuel offers more reassurance in a crisis than pipeline-supplied natural gas, which could be disrupted by an earthquake, flood or other extreme event.

Despite their robust natures and practical attributes, however, diesel-engine generators require certain levels of care and predictive/preventive maintenance—especially units that function in a standby capacity. Discovering in the middle of a power blackout that a site’s critical standby-power system won’t run is a high-stress headache no maintenance department needs. Allowing water to enter undetected and wreck havoc in these units’ lubrication systems is a sure way to induce that type of headache.

Cooling-water intrusion into the lubrication system often results from a faulty gasket around a cylinder head, a cracked cylinder liner, a warped head or uneven bolting of the cylinder head to the engine frame. In any case, water in the engine oil is undesirable. Such intrusion can have a significant impact on the unit’s ability to run: If enough moisture has entered the lubrication system, water carried by the oil may evaporate during the combustion cycle and leave dry spots that allow the piston and cylinder walls to make metal-to-metal contact. This usually results in damage to the cylinder, the liner and the engine. Continued operation of equipment with this condition will damage the unit and can lead to complete engine failure. With that in mind, consider the following example from the real world.

Uncovering the problem

In a routine test run of a standby 5000 KW, 16-cylinder diesel-engine generator unit, water was observed in a lubrication-oil sight glass. Subsequent investigation discovered a jacket-water leak in the 1-left cylinder.

Fig. 1. Leakage observed in 1-left cylinder in situ with bore scope

During troubleshooting, the water jacket was pressurized to 11 psig, and water began running off the piston on the interior of the cylinder liner. The fuel injector was then removed and a bore scope employed to examine the internal area. As shown in Fig. 1, the liner was found to be leaking from a point about one inch below the piston-ring reversal area and filling the top of the piston. Subsequent examination indicated that the liner had cracked.

As with many units, the water jacket had an automatic refill feature. When the coolant level in the engine-water jacket drops, cooling water is automatically replaced in the standpipe by a level-control device. Given this arrangement, if a leak were to occur in the water jacket, any “missing” water would not be noticed. Consequently, attempting to determine when the liner crack started by checking for “missing” coolant was a dead end.

The coolant consisted of demineralized water with an added corrosion inhibitor: sodium nitrite. Since coolant leaking into the lubrication system would also carry with it sodium nitrite in solution, the presence of this corrosion inhibitor in the oil was used as a marker to indicate when cracking in the liner had sufficiently developed to allow coolant intrusion into the oil. The engine oil was regularly tested, and one of the tests looked for sodium content.

A review of oil-analysis reports showed that on the day of the engine test during which water was observed in the sight glass, the sodium content of the oil exceeded 15 parts per million (ppm)—which was the trending alert level. The sodium content of the oil versus time, as reported in oil analyses, was then plotted as shown in Fig. 2. Since the oil was sampled at quarterly intervals, the resulting plot appears choppy. The plot in Fig. 2 is further complicated by the fact that the oil had been changed several times during the period of the plot. Each oil change would, of course, knock the concentration of sodium back to zero. Despite this complication, interpretation of the plot in Fig. 2 suggests that leakage had likely begun two years earlier.

Since the concentration of the sodium nitrate in the jacket-cooling water and the sodium content in the oil were known, it was possible to estimate the amount of water that had crossed into the lubrication system. This was then compared to the amount of water that had layered out in the bottom of the lubrication reservoir and that remaining in solution in the oil.

When these evaluations were done, the estimate of water carryover through the liner crack matched the total amount of water that had both layered out and remained in solution. This finding further corroborated that the origin of the sodium was the jacket water—and not any other source. The use of the sodium concentration data depicted in Fig. 2 also allowed rough estimates of leakage rates when engine operating time and oil replacements were considered.   

While a laboratory report for an oil sample taken prior to the engine-test run had indicated a sodium content of 18 ppm, it also indicated nil for water content. Water content for the previous six summary laboratory reports also indicated nil with respect to water content. These findings begged the question: If water had been leaking into the lubrication oil long before the test run, why didn’t the previous laboratory reports pick it up? Answer: Because the previous reports had indicated “nil” water, the significance of the sodium noted in previous reports was overlooked. To better understand this chain of events, keep in mind that water can be present in lubrication oil in one, two or three forms: 1) in solution; 2) in an emulsion; 3) as free water.

If not otherwise already saturated, oil is able to take a certain amount of water into solution. In this condition, it will look clear and cause minimal problems in lubricated equipment. If, however, oil has taken into solution all the water it can—i.e., when it is saturated—any excess moisture is held in suspension or emulsion. In this condition, it looks cloudy. If more water is added, the oil and water will separate into two layers. Because water is heavier than most lubricating oils, excess water will sink to the bottom. This is “free” water.

To get a sense of the amounts of water associated with these three forms of water, consider the following:

• A saturation level of a mineral oil might be about 100 ppm, and the amount of water it can hold in emulsified form could be as much as 1000 ppm. Free water occurs when the concentration is greater than 1000 ppm.
• Similarly, an ester-based hydraulic fluid could have a saturation level of more than 2000 ppm, and also might be able to hold as much as 5000 ppm in emulsified form. Water in excess of 5000 ppm would then form a free water layer under the oil.

The amount of water that can be taken into solution also depends on oil temperature (as shown by the graph in Fig. 3). Oil at a higher temperature generally can take more water into solution than oil at a lower temperature. (The lubrication oil in the referenced diesel-engine generator unit is maintained at a “ready to run” temperature no lower than 104 F and no higher than 185 F.)

In this case, the water content of the lubrication oil was checked using the “Crackle Test.” This procedure involves putting two drops of oil on a hot, flat surface (about 130 C) and documenting any bubbling or “crackling” (the sound a sample sometimes makes during this type of test). The size of the bubbles is then used to indicate the amount of water in the oil. For example:

• If no change is observed in the two drops of oil on the hot surface, it can be concluded that no free or emulsified water is present.
• If very small bubbles (about 0.5 mm in diameter) appear, then disappear, the water content can be estimated at 500 to 1000 ppm.
• If bubbles of about 2 mm appear, move to the center of the hot plate, increase in size to 4 mm, then disappear, the water content can be estimated at 1000 to 2000 ppm.
• If bubbles of about 2 to 3 mm appear, increase in size to 4 mm and the process repeats with possible violent bubbling and audible cracking (hence, the test’s name), the water content is more than 2000 ppm.
• The following limitations to the Crackle Test, however, are crucial to the discussion of this case study:
• The method is approximate and not considered quantitative. It simply provides a rough indication.
• If the hot-plate temperature is above 130 C, the heat can induce rapid scintillation that may not be detected by the observer.
• The method provides no information about the amount of water in solution in the oil.

Developing new best practices

Based on the limitations of the Crackle Test and the data in Fig. 3, it became clear to the site’s maintenance department that water content could “hide out” in solution within the oil of a generator’s lubrication system. This situation had gone undetected over time because oil samples were usually taken right after a test run, when the lubrication oil was still hot (higher temperature => more water in solution).

In short, to proactively detect engine-cooling-water intrusion into the lubrication oil system before it affects the readiness of this site’s standby-power equipment, the amount of oil in solution must be monitored along with the amount of sodium detected in the oil.

Furthermore, personnel now understand that before the standby unit’s lubrication oil is changed—or fresh oil is added—samples of the existing oil must be evaluated. This way, any increase in sodium or water in solution can be compared to where it left off the last time.

Randy Noon is a Root Cause Team Leader at Nebraska’s Cooper Nuclear Station. A Registered Professional Engineer, book author and frequent contributor to Maintenance Technology, he has been investigating failures for more than three decades. Contact him at rknoon@nppd.com.