Fault-finding on board is a measurement discipline: this chapter works through why insulation resistance is read as a trend, how an earth fault is traced without guesswork, and how condition monitoring gives you weeks of warning before something trips.
An insulation resistance test pushes a DC test voltage across the gap between a winding or cable conductor and earth and measures how much current leaks through. In new, dry insulation that leakage path is enormous, so the meter reads high — hundreds of megohms or more. As insulation ages, absorbs moisture, or collects conductive dust and oil film, extra leakage paths open up in parallel with the insulation itself, and the reading falls. That is the mechanism: you are not measuring a material property so much as counting how many stray conduction paths have formed.
A single reading on its own tells you almost nothing, because "how much resistance is enough" depends on the machine, its size, its age and the conditions on the day. What tells you something is the trend: the same motor's reading falling from 400 MΩ to 40 MΩ over eighteen months, even though 40 MΩ still clears the ≥1 MΩ rule-of-thumb threshold, is the early warning a single spot reading would miss entirely. Record temperature and relative humidity with every test, because both affect the reading independently of the insulation's actual condition — warm, humid air lowers a reading on perfectly sound insulation.
The polarisation index gets round the "no baseline" problem for a first-ever test. As the DC test voltage is applied, a genuinely dry, sound insulation's resistance keeps climbing for the first several minutes as internal polarisation currents die away; a contaminated or wet insulation's resistance has already levelled off by one minute, because there is a continuous conduction path doing the work instead. A PI at or above 2 indicates healthy, dry insulation; a PI below 1 is a clear warning regardless of what the raw megohm figure says.
Treat every IR figure as one point on a graph, not a pass/fail number — and never compare two points taken at different temperatures without correcting one of them first.
Most ship's distribution is an insulated (IT) system by design: no point of the network is solidly connected to the hull, so a single conductor touching earth does not, by itself, create a fault current path back to the source. That is deliberate — it means one earth fault does not have to trip anything, and the vessel can keep running while the fault is found and cleared in an orderly way. The earth fault lamps or the insulation monitoring device exist purely to tell you that this first fault has happened; they do not tell you where.
Locating it is a process of elimination, not guesswork. Starting from the section the monitor points to, breakers are opened one at a time and the indication is watched: when it clears, the fault sits on the circuit just isolated. The order matters as much as the method — before opening any breaker, weigh what that load is and what losing it does to the vessel, and open the least essential circuit first. A fault found by blacking out the switchboard in the wrong order is a worse outcome than the original fault.
Finding and clearing that first fault is not optional housekeeping — it is what keeps the system insulated. A second, independent fault on a different phase, while the first is still live on the system, connects two phases together through two separate earth points. That is no longer a monitored, non-tripping condition; it is a phase-to-phase fault carrying full short-circuit current through whatever provided the earth paths, with all the arcing and fire risk that implies. The insulated system's entire safety case rests on there never being two faults present at once, which is why "nothing tripped" is not a reason to leave a single fault alone.
An insulated system's first fault is a warning you are given time to act on. That time is the entire benefit of the design — it is not a reason to wait.
Faced with a fault that is not obviously visible — a control circuit that will not energise, an alarm that will not clear, a pump that trips on start — the temptation is to start changing parts: relay, then contactor, then cable, until something works. That approach is expensive, slow, and worst of all it can "fix" the symptom by coincidence — a connector reseated while a relay was swapped — while leaving the real cause, often a marginal connection or a chafed cable, exactly where it was to fail again.
The disciplined alternative is a half-split search. First confirm the symptom is real and repeatable — operate the circuit again and watch exactly what does and does not happen, because a fault described from memory is often not quite the fault that is actually present. Next, establish what has changed: recent maintenance, a modification, a period of heavy weather, water ingress, or simply nothing identifiable, which is itself useful information. Then define the boundaries of the suspect circuit from source to load, and test at the electrical midpoint — a terminal, a junction box, a relay coil — rather than at either end.
Whichever half of the circuit shows the fault, discard the other half from suspicion entirely and repeat the split within the remaining half. Each test roughly halves the amount of circuit left to search, so a fault in a run of a dozen components is typically found in three or four well-chosen tests rather than a dozen ad-hoc ones. The method works on wiring, on control elements within a piping system, and on any system that can be meaningfully divided — its value is that it converges on the actual point of failure with evidence at each step, rather than on a guess that happened to work.
Every test should eliminate half of what remains under suspicion. If a test does not narrow the search, it was the wrong test to run.
Four instruments cover almost all fault-finding on board. A multimeter gives voltage, continuity and low-range resistance for everyday checks. An insulation resistance tester (megger) applies a DC test voltage — commonly 500 V for low-voltage circuits and up to 5 kV for high-voltage plant — well above normal working voltage, specifically to stress the insulation and reveal a weakening path that a low-voltage continuity check would never find. An earth fault or insulation monitoring device watches the whole insulated system continuously rather than at the moment of a spot test. A clamp meter measures current without breaking the circuit, useful for confirming an unbalanced load or a running current that does not match the nameplate.
Because the insulation tester deliberately applies a high test voltage, it must never be connected to a circuit with electronics still in it — control cards, VFDs, sensors and similar components are built for their working voltage, and the test voltage will punch through them. Isolate the circuit, disconnect or bridge out any electronic device on it, and only then test. A long run of cable also behaves as a capacitor: the test charges it, and that charge must be discharged to earth after the test, before anyone touches the conductors or continuity checks follow.
The standard order for testing a circuit that has just been repaired or reinstated is continuity, then insulation, then polarity, then function — in that sequence, and not skipped. Continuity first confirms the conductors are actually joined the way the drawing says; insulation next confirms there is no unintended path to earth or between cores; polarity confirms supply and neutral, or the correct phase sequence, have not been swapped; only once all three are satisfied is it safe to energise and functionally test the circuit.
Preventive maintenance replaces or services a component on a fixed interval regardless of its actual condition; condition monitoring instead measures a physical quantity that changes as a fault develops, so the component can be worked on when it actually needs it — often catching a developing fault weeks before it would otherwise trip or fail outright. Three techniques cover most rotating and electrical plant on board, and each is sensitive to a different failure mechanism.
Thermography — an infrared camera survey of switchboards, terminations and connections — finds loose or high-resistance joints, because a poor connection has higher resistance than a good one and therefore dissipates more heat under the same load current; it shows up as a hot spot well before that joint fails or arcs. Vibration analysis on rotating machinery reads the frequency spectrum against the shaft's running speed: a dominant peak at once-per-revolution (1×) points to unbalance, a dominant peak at twice-per-revolution (2×) points to misalignment, and defect frequencies higher again typically point into the bearings themselves. Insulation resistance trending, covered earlier, is the third stream, watching the winding's condition over time rather than at an instant.
All three depend on the same thing: a baseline. A single thermal image, a single vibration reading or a single IR figure describes a moment, not a trend, and without a reading taken at a known-good condition and a comparable load there is nothing to measure the new reading against. That is also why the paperwork matters as much as the test — every entry should record as-found condition, the action taken, and as-left condition, because that record is what turns this month's spot reading into next year's baseline, and it is the evidence, on inspection, that the maintenance was actually done and why.
A condition-monitoring reading is only as useful as the baseline it is compared against — take the baseline before you need it, not after something has already gone wrong.
The three examples below apply the same reasoning under exam conditions: read the given data carefully, decide which rule applies, and show the working that gets you to the figure — not just the figure itself.
An insulation resistance test is carried out on a 440 V motor during the ship's annual survey. At 1 minute the meter reads 20 MΩ; at 10 minutes it reads 50 MΩ. The motor's previous IR test, six months ago, recorded 80 MΩ at an ambient temperature of 20 °C; today's test was carried out at 30 °C. Using the rule of thumb that insulation resistance halves for every 10 °C rise in temperature, assess whether the motor's insulation is healthy and whether the trend is improving or worsening.
IR at 1 min = 20 MΩ IR at 10 min = 50 MΩ Previous test: 80 MΩ at 20 °C (6 months ago) Today's test temperature: 30 °C Rule of thumb: IR halves per 10 °C rise
Check the polarisation index first.
It needs no history and tells you whether the insulation is dry.
A PI of 2.5 is above the ≥2 threshold.
So the insulation is dry and in good condition at the moment of test. Next compare today's reading against the last recorded value — but the two tests were taken 10 °C apart, so the raw figures cannot be compared directly.
Compare like with like.
With both readings referred to 20 °C, the trend across six months can be read fairly.
AnswerPI = 2.5 (healthy, dry insulation); the temperature-corrected reading has risen from 80 MΩ to 100 MΩ over six months — the trend is stable to improving, not deteriorating.
The trap: comparing the raw 50 MΩ reading directly against the previous 80 MΩ without correcting for the 10 °C temperature difference would wrongly suggest the insulation has degraded, when in fact it has not.
After the insulation monitor on the main switchboard indicates an earth fault, breakers are opened one at a time in order of least operational consequence until the indication clears on opening the No. 2 generator cooling-pump motor feeder breaker. With the circuit isolated at both ends except for the faulted core, the resistance from the switchboard terminal to earth (the hull) is measured as 0.51 Ω. The cable manufacturer's data sheet gives a core resistance of 3.0 Ω per km at 20 °C. The cable route has junction boxes at 60 m and 130 m from the switchboard, and terminates at the motor at 210 m. Estimate where the fault lies and identify which section of the run should be opened up first.
Measured resistance, switchboard terminal to earth via faulted core = 0.51 Ω Cable core resistance = 3.0 Ω/km at 20 °C Junction box A at 60 m; junction box B at 130 m; motor terminal at 210 m
Estimate where the fault lies and identify which section of the run should be opened up first
Confirm the fault has been isolated to one circuit.
Opening breakers one at a time and watching the indication clear has already narrowed the fault to this single feeder — the resistance measurement now locates it along that one known cable run.
Convert the measured resistance to a distance using the cable's resistance per unit length; up to the fault, the faulted core behaves as an ordinary length of conductor to earth.
Locate the figure on the actual route.
Compare the calculated 170 m against the known positions of the junction boxes.
AnswerThe fault is approximately 170 m from the switchboard, in the cable run between junction box B (130 m) and the motor terminal box (210 m) — open that section first, starting with the motor gland and terminal box.
The trap: this calculation assumes the measured resistance path runs only through the single faulted core to the one earth point found; an unnoticed second earth contact elsewhere on the system provides a parallel path that lowers the reading and makes the calculated distance shorter than the true fault position.
A seawater cooling pump motor runs at a nameplate speed of 2880 rpm. At commissioning, the baseline vibration spectrum showed an overall reading of 1.2 mm/s, dominated by a peak at the running-speed (1×) frequency, with no significant second-order peak. Monthly condition-monitoring readings over the following four months show overall vibration of 1.2, 1.3, 2.6 and 4.1 mm/s, and by the fourth reading the spectrum's dominant peak has shifted to twice running speed (2×). The vessel's planned maintenance system sets an alarm limit of 4.5 mm/s overall. Diagnose the likely developing fault and decide whether to act now or wait for the alarm.
Motor speed = 2880 rpm Baseline: 1.2 mm/s, dominant peak at 1× Monthly overall readings: 1.2, 1.3, 2.6, 4.1 mm/s Latest spectrum: dominant peak at 2× running speed PMS alarm limit: 4.5 mm/s overall
Identify the running-speed frequency so the spectrum's peaks can be read against it.
The dominant peak has moved from 1× at the baseline to 2× now.
Per the 1×/misalignment, 2×/unbalance rule, that shift points to a specific developing mechanism rather than generic wear.
Check the rate of change.
Not just the latest figure, against the alarm limit.
AnswerThe 2× dominant peak indicates developing misalignment, not unbalance; the trend is projected to cross the 4.5 mm/s alarm within about a month. Schedule a shaft alignment check at the next opportunity rather than running to the trip.
The trap: waiting for the overall figure to cross the hard alarm before acting — the frequency signature already identifies the mechanism and the rate of change already gives the timescale, and that early warning is the entire point of condition monitoring.
IR ≥ 1 MΩ (rule of thumb)Trend across readings matters more than a single valuePI = R₁₀ / R₁ ≥ 2Healthy, dry insulation; PI < 1 suggests contamination or moistureTest voltage: 500 V (LV) / 5 kV (HV)Disconnect electronics first; discharge the cable afterwardsIR halves per ~10 °C rise (rule of thumb)Correct readings to a common temperature before comparingEarth fault: open breakers one at a timeWeigh the operational consequence of each load before opening itContinuity → insulation → polarity → functionThe standard sequence for testing a repaired circuitHalf-split (binary search)Test at the midpoint of the suspect section, then halve againThermographyLoose or high-resistance connections run hot under load before they failVibration: 1× unbalance, 2× misalignmentHigher-order defect frequencies point into the bearingsRecord: as-found / action / as-leftThe PMS entry is the evidence the work was done and why