Keeping oil pressure above water pressure can favor oil flowing outward at a single separating-wall leak. That preference is a contamination-control measure, not proof that water cannot enter oil. Pump trips, hydrostatic head and pressure transients can reverse the local difference. [1]
Compare pressures at the same place
Define Δp(t) = poil,local(t) − pwater,local(t) in pascals. Readings from gauges at different elevations require a static-head correction: pressure rises by ρgΔz when moving downward through a stationary fluid. Use density ρ in kg/m³, gravitational acceleration g in m/s² and vertical distance in metres.
During flow, add the appropriate friction and local pressure changes between gauge and cooler. A pump discharge pressure does not directly establish pressure on the far end of a tube bundle.
Build the envelope in states, not averages
Normal running: record both flows and local inlet/outlet pressures.
Oil pump stopped, water available: assess whether water head remains while oil pressure falls.
Water pump stopped: include connected supply head and any siphon or isolation condition.
Both pumps stopped: calculate trapped-fluid and static pressures from the actual elevations.
Valve operation and pump trips: evaluate transient peaks and any reversal of Δp.
Use actual controls and isolation assumptions. A state requiring a closed valve is not credible if the plant logic leaves that valve open.
A cooling-system review should verify the real valve state and oil circulation; maintenance guidance documents both misleading valve indication and low oil flow [2]. Use that evidence to establish which pressure boundaries actually remain connected after a trip.
Correct both pressure readings to the cooler
Assume an oil gauge at cooler elevation reads 130 kPa gauge. A water gauge 5 m above the same cooler point reads 100 kPa gauge. With assumed water density 1,000 kg/m³, g = 9.81 m/s² and stationary columns, cooler water pressure is 100 + 1,000 × 9.81 × 5/1,000 = 149.05 kPa gauge. Thus local oil-minus-water pressure is −19.05 kPa despite the apparent +30 kPa gauge margin. Repeat the correction separately for both fluids in every connected operating state, then include flowing losses and transients.
Two acceptance questions remain separate
The first question is whether the pressure difference supports the intended leak direction in the specified states. The second is whether each circuit and separating feature remain within their permitted differential and absolute pressures, including transients.
References
- U.S. Department of Energy, DOE-HDBK-1012/3-92, Thermodynamics, Heat Transfer, and Fluid Flow, Volume 3, 1992.
- U.S. Bureau of Reclamation, FIST Volume 3-30, Transformer Maintenance, 2000.