RECD Service Visit Found No Fault: Planning Better Intermittent-Fault Records

The service report is tidy. Date, engineer’s name, a list of checks, all marked “OK”, and at the bottom a line that nobody on the site wanted to read: no fault found. The technician ran the set, watched the RECD for an hour, saw nothing unusual and left. Two days later the alarm came back.

It is easy to be angry at the technician, and sometimes that anger is justified. More often the report is perfectly honest. The fault did not appear during the visit, so it could not be seen. That is the nature of an intermittent fault: it behaves well when someone is watching. The usual response, a second visit with the same approach, tends to produce the same report. What changes the outcome is not a better technician on the day but better records before the technician arrives, a visit that is planned to catch the fault, and an agreed way of deciding when a “no fault found” can be treated as closed.

This article is about that planning. We look at what “no fault found” really means, set out a complaint card that captures the conditions around a fault, describe a capture ladder of four levels of recording, explain how to brief the technician and what to leave behind, and finish with the conditions that should be met before an intermittent fault is called resolved. The advice is general. Your supplier’s service procedures and your service contract take priority.

What “no fault found” really means

The phrase hides several different situations, and treating them as one leads to wasted visits.

1. The fault was not present. It came and went, and the visit fell in a quiet spell. The most common situation.

2. The fault was present but the test did not reproduce the condition. The visit tested at 30% load in dry weather and the fault needs a start from cold after a rain, or a load step, or a certain regeneration phase.

3. The fault was in a system the technician did not examine. The visit looked at the RECD but the cause lies in the DG controller, the ATS, the supply or a sensor wiring run.

4. The records were not good enough to know. The technician did not have the alarm times, the values or the conditions, so no test could be aimed.

A reference article on intermittent faults explains why they are so hard: such a fault tends to result from several contributing factors, some of them effectively random, occurring together, and each factor on its own does not create the problem, so the factors can only be identified while the malfunction is actually happening. Logging the relevant parameters over a long enough time can show the values at the time of the fault (Wikipedia: intermittent fault). An academic paper on intermittent fault strategies calls them the most difficult class of faults to diagnose and a main root cause of no-fault-found outcomes (Coventry University).

An automation-industry troubleshooting article adds a human point: technicians are often under pressure simply to get the equipment running again and not to find the root cause, and good questioning of those who observed the fault, about what they were doing and what the system was doing, is among the most valuable steps (Automate.org). That is a polite way of saying that the quality of the first report from the site decides the quality of the diagnosis.

So when a visit ends in “no fault found”, the useful question is not “who failed?” but “which of the four situations are we in, and what record would tell us?”

Before the visit: the complaint card

A complaint card is a one-page record, filled in by whoever sees the fault, at the time it happens. It is the single most useful document in an intermittent-fault investigation. It should be quick to fill in, which means mostly ticks and numbers, with a few words of description.

Fields for the card

FieldWhat to write
Date, time (to the minute), set ID, who observed
What happenedAlarm text or code; trip; smoke; noise; reading out of range. Photograph the screen.
Was the alarm still active when you looked?Active / cleared by itself / acknowledged
Set stateLoad (kW or %), time since start, start type (cold, warm, hot restart), load step in the last 10 minutes
Values on screenΔP, temperatures, heater current, supply voltage, anything shown
ConditionsWeather in the last 24 hours, plant-room temperature, outages, other equipment starting
Recent changesService, fuel delivery, setting change, repair, new load
What was doneAcknowledged, reset (which type), restarted, called someone
What happened afterwardsAlarm returned after N minutes or did not return
Photographs and attachmentsScreen, exhaust outlet, stack, sensing hardware

Keep a pad of these at the controller, or a form on a phone, and tell shift staff why they matter. Our guide to the RECD shift handover log shows how to pass the information on, and what to record in an RECD event log shows a longer-term format.

The conditions diary

Intermittent faults are often tied to conditions nobody was thinking about. A simple diary kept beside the complaint cards helps. Each day, note the date, run hours, number of starts, any outage, rain, extreme heat, new equipment, and anything unusual. After a few weeks, patterns that nobody had expected tend to show up. Our guide on RECD faults that appear only in wet weather is one example of a hidden condition, and RECD pressure spikes only at startup is another.

The capture ladder: four levels of recording

Not every site can install a data logger tomorrow. The capture ladder describes four levels, each of which adds something. Start at the first rung and climb as the problem demands.

Level 1: human notes with photographs

The complaint card, the diary and photographs. Costs nothing, can be started today and catches the context that no instrument can. Limitation: it captures what a person happens to see, and short events go unnoticed.

Level 2: controller logs and exports

Most RECD and genset controllers keep an alarm history, an event log and sometimes trend data. Learn how to export these, and export them regularly, before any reset or visit. Photograph the history screen if export is not possible. Check the logging interval. A log that records a value once a minute cannot show an event that lasts five seconds. Check the clock too: if the RECD controller, the DG controller and the building system disagree by minutes, their logs cannot be lined up. See aligning RECD sensor timestamps with generator runtime logs and detecting missing data in RECD monitoring records.

Level 3: an independent logger

An independent logger records a few channels at its own sampling rate, away from the controller. For an RECD, useful channels are ΔP from a separate transmitter, a temperature or two, controller supply voltage and, if you can arrange it, load current. An independent record answers a question the controller cannot answer about itself: was the alarm real, or was the controller’s own sensing the problem? Automotive diagnostic writers describe the same approach: put recording equipment on the system and send it back into service with instructions to return when the fault occurs, and review the data afterwards. One technical note on diagnosing intermittent faults describes a logger that records many values over hours, enabling a fault that appeared only after five hours of use to be found in the data (TEXA technical note).

Level 4: event-triggered capture

The most powerful level. The logger records continuously into a short buffer and, when a trigger occurs (an alarm output, a threshold crossing), saves the data from before and after it. The automation-industry article above describes the idea: let the system run for hours, and when the fault occurs, capture a short window of key data such as supply voltage, current and speed from just before the event. This catches fast events that slow logging would miss. It needs a logger with a trigger input and a pre-trigger buffer, and a little planning of which signal triggers it.

Choosing the level

Fault characterMinimum level that is likely to help
Recurs daily or weekly, with a visible patternLevel 1 plus Level 2
Recurs rarely, tied to conditionsLevel 2, with exports after every event
Reading disagrees with expectation; sensor suspectedLevel 3
Lasts seconds, no obvious triggerLevel 4
Affects safety or complianceGo to Level 3 or 4 early

A small recording plan

For a Level 3 or 4 set-up, agree beforehand what to record, at what rate and for how long.

ChannelSuggested rateWhy
ΔP across the RECD (independent transmitter)1 sample per second, or faster for transientsShows whether the alarm was real
Back-pressure at engine outlet (if tapped)1 per secondEngine limit
Inlet and outlet gas temperature1 per second to 1 per 5 secondsSlow-changing, but needed for regeneration
Controller supply voltage10 per second or fasterBrownouts are brief
Heater current1 per secondConfirms heater action
Load (kW)1 per secondEverything depends on it
Alarm digital outputsEvent-basedTime-stamps the trigger

These figures are starting points. Ask the supplier which channels they would want and what rate suits the device.

Briefing the technician

A visit is a short, expensive opportunity. Prepare for it as you would for an inspection.

Send a pack ahead of time.

  • The complaint cards and the conditions diary.
  • Controller exports and photographs.
  • The timeline of events (see RECD faults after a fuel delivery for the timeline method).
  • A list of recent changes: service, repairs, deliveries, settings.
  • The question you want answered, in one sentence: “Why does the high-ΔP alarm appear on cold mornings?”

Ask for the conditions to be reproduced. If the fault happens on cold starts, ask the technician to be present for a cold start. If it happens at load steps, arrange a load step. If it happens in damp weather, ask whether the visit can be timed for a damp morning, or whether a controlled moisture check is possible. A visit at noon on a dry day is the wrong test for a morning fault.

Ask for a baseline while it is working. An automotive diagnostics columnist advises capturing data and waveforms while the circuit is operating properly, so that there is a known-good pattern to compare against later (MOTOR). A baseline taken during the visit, with a good record of ΔP, temperatures, heater current and supply voltage at three loads, is the most valuable thing a “no fault found” visit can leave behind.

Agree the no-reset rule. Ask that history, counters and learned values not be cleared during the visit unless the ledger is complete. See RECD alarms that return after a reset.

Agree who else needs to be present. The cause may sit with the DG engine team, the electrical contractor or the building system owner. Our guide on coordinating DG engine repairs with the RECD service team explains how to share the job.

Check the contract. Response times, repeat-visit terms and what counts as a closed job should be in the agreement. See RECD service response commitments and third-party service companies maintaining an RECD.

During the visit: what to ask for

Technicians welcome a customer who knows what to ask for. During and at the end of the visit, request the following.

  1. A record of what was tested and under what conditions. Load, duration, start type, weather and whether the fault condition was reproduced.
  2. A list of what was ruled out. “Sensor line checked and dry, heater current normal at 50% load, supply voltage 24.1 V under cranking” is far more useful than “all OK”. Our guide to auditing an RECD service report for missing information lists what a report should contain.
  3. The baseline snapshot described above.
  4. A recorded reason if the fault could not be reproduced: which conditions were absent.
  5. A recommendation for the next step. A logger left on site, a second visit under specified conditions, a supplier engineering review or a precautionary replacement.
  6. Agreement on what the logger will do if one is left: channels, rate, trigger, who retrieves the data and when.

If the technician cannot stay for a cold start at 6 a.m., record that. The report should say so, and the gap becomes part of the evidence.

After the visit: closing a no-fault-found report

A no-fault-found report is not a closure. It is a status. Agree in advance what it takes to close an intermittent fault, and what happens if the fault returns.

Suggested closure conditions

  • A defined observation period, expressed in run hours and in conditions, not only in days. For a cold-start fault, for example, at least N cold starts across at least M different mornings, including a damp one.
  • No recurrence in that period, with the complaint cards and logger data confirming that the conditions did occur.
  • A baseline on file for comparison.
  • The ledger and cards signed off by someone other than the person who wrote them.

Escalation triggers

  • Three occurrences without a diagnosed cause trigger a formal fault-isolation procedure, not another reset. An aircraft manufacturer’s maintenance guidance takes exactly this line: after three occurrences of a failure, even if it has not been confirmed, apply the fault-isolation steps in the manual until it is resolved, because resets can be a first step but should not be the last (Airbus).
  • Any occurrence connected with smoke, trip, high temperature or a compliance risk escalates at once.
  • A logger capture that shows a real event moves the case to the supplier’s engineering team.

Precautionary changes. Sometimes replacing a suspect part, a sensor, a tube or a connector, is a reasonable gamble, particularly when it is cheap and the fault matters. The reference article on intermittent faults notes that precautionary changes without pinpointing the fault are one of the techniques used, and the obvious drawback is that you may never learn what was wrong. If you do it, record what was changed and the date, so that a return of the fault is informative. A fault that returns after a sensor swap tells you the sensor was not the (only) cause.

Who owns the evidence

Intermittent fault records fail most often because nobody owns them. Name an owner for each of the following.

ItemSuggested owner
Complaint cards and diaryShift supervisor
Controller exports and clock checksMaintenance engineer
Independent logger and its dataMaintenance engineer, with supplier input
Visit briefing packMaintenance lead
Closure decisionMaintenance lead and facility manager together
Contract and response termsProcurement or facility manager

A named owner, a monthly check that the records are being kept and a rule that nobody clears history without a card are enough to change outcomes.

A worked illustration: the Tuesday mornings

The following is an illustrative scenario, not a real customer. A school campus has a 250 kVA set with an RECD that raises a high-pressure alarm roughly once a fortnight. Two service visits, both in the afternoon, ended in no fault found. The facility manager decides to change the approach.

Level 1. Complaint cards are started. After six weeks, six alarms have been recorded. Five of them were on Tuesday or Wednesday mornings, at the first start after the weekend test run, and three of those came after rain.

Level 2. The controller exports show alarms about 40 seconds after start. The clocks on the RECD and DG controllers differ by about four minutes, which is corrected.

Level 3 and 4. The supplier leaves an independent differential-pressure logger with a trigger on the alarm output, recording a one-second sample and a ten-minute pre-trigger buffer. On the next Tuesday alarm, the capture shows the controller’s pressure reading spiking while the independent transmitter shows only a gentle rise tracking load.

The visit brief. The supplier is asked to visit on a Tuesday morning, after a damp night, at the first start. The technician finds a sag in the high-pressure sensing tube holding a small amount of water.

Closure. The tube is re-routed. The team sets a closure condition of eight cold starts across four mornings, including at least one after rain, with the logger left in place. No alarm occurs. The case is closed, with the baseline, the logger data and the cards filed. The earlier two “no fault found” visits were correct reports of a fault that only appeared on a cold, damp start.

A one-page visit brief to adapt

Complaint: (one sentence)
When it happens: (conditions, from cards and diary)
What we have: (cards, exports, logger data, photographs)
What we need from the visit: reproduction of the conditions, baseline at three loads, independent check of the sensing chain, list of items ruled out
Conditions to be present: (start type, load step, weather, time of day)
No-reset rule: history and learned values not to be cleared without notice
Who else attends: (engine team, electrician, building system owner)
Output: written list of tests and conditions, baseline snapshot, next-step recommendation, logger plan
Closure condition: (observation period and occurrences)

Questions people ask

Is “no fault found” a sign of a poor service?

Not necessarily. It is an honest statement that the fault did not appear. The question is whether the visit was planned to catch it, and whether the report lists what was ruled out.

How long should I keep logging?

Long enough to catch several occurrences and the conditions around them. For a fault that recurs fortnightly, that means at least a few cycles, often six to eight weeks.

Is an independent logger worth the cost?

For a fault that matters, a recurring compliance risk or a repeated unplanned stop, it usually pays for itself quickly, because it separates a real exhaust event from a sensor or controller problem.

Can the controller do the logging by itself?

Sometimes. Check what it stores, how often, for how long and whether you can export it. If the controller is the suspect, an independent logger is a better witness.

When should the supplier’s engineering team get involved?

When a capture shows a real event you cannot explain, after three occurrences without a diagnosis, or whenever safety or compliance is affected.

In short

An intermittent RECD fault will not appear to order, so the answer is to catch it. Keep a complaint card and a conditions diary, climb the capture ladder, brief the technician with a pack that tells them which conditions to reproduce, ask for a baseline and a list of what was ruled out, and agree in advance what it takes to close the case. If you would like help planning a logger set-up or a visit brief, you can browse our retrofit emission control devices, read the FAQ or contact the Aceget team.

Related Reading

Sources



Leave a Reply