At a glance
| Operator | Excursion railroad, Colorado |
| Application | Generator car pulled behind the locomotive, powering HVAC and appliances for the train |
| Configuration | Two onboard generators, MTU engines, one running and one held as a spare |
| Reported problem | One unit down, suspected loss of communication between the genset control module and the engine ECM |
| Second unit | Also reported as having issues |
| Operating schedule | Seven days a week, two to three runs per day |
| Service access | Only in the windows between runs |
| Stopgap in place | A portable generator on the back of the train to finish the day |
| Call received | Emergency service |
The call
Sunday morning, emergency line. An excursion railroad with a generator car behind the locomotive, two MTU sets in it, one carrying the load and one held as a spare.
The primary was down. The mechanic had looked at it Saturday, called it a communication fault between the control module and the engine ECM, and was unreachable for the weekend. The spare had open issues of its own. The train was running anyway, on an older portable strapped to the back end.
They wanted a tech the next day, in whatever gap the run schedule left.
Three problems in that, and all three are common enough to be worth going through one at a time.
What a communication fault between the module and the ECM actually is
Start with the thing most operators do not realize about a modern genset: there are two computers in it, and they are not the same computer.
The engine has an ECM, the engine control module, supplied by the engine manufacturer. It runs fuel delivery, timing, and engine protection, and it owns the sensors bolted to the engine itself. Oil pressure, coolant temperature, speed, fuel rail.
The generator set has its own controller, which is the box with the display on the side of the machine. It handles starting and stopping, voltage regulation, breaker control, transfer signaling, alarms, and the operator interface.
Those two talk to each other over a data link, almost always a CAN bus running the SAE J1939 protocol. The controller does not read the engine’s sensors directly. It asks the ECM, and the ECM answers, dozens of times a second.
When that link drops, the controller has no idea what the engine is doing. What you see depends on the platform, but the common symptoms are recognizable:
A fault code specifically calling out lost communication or a data link failure, rather than an engine fault. Gauges on the controller reading zero or dashes while the engine itself is mechanically fine. A unit that cranks and will not run, or that starts and then shuts down a few seconds later on a protection fault that never actually happened. A unit that will not respond to a start signal at all.
The important part for anyone standing in front of one of these: a communication fault usually is not an engine problem. The engine is often perfectly healthy. What has failed is the conversation about the engine. That distinction matters because it changes what gets inspected first and it usually means the repair is smaller than the symptom suggests.
The usual causes, roughly in the order a tech works through them:
Connectors first, because they are the most common and the cheapest to fix. Corrosion, a pushed-back pin, a latch that vibrated loose. Then harness damage, chafe points, and anywhere the loom passes near something hot or something that moves. Then the CAN terminating resistors, which have to be present at both ends of the bus and which fail or get disturbed during other work. Then power and ground. A weak battery, a marginal charging circuit, or a corroded ground strap will produce communication errors that look nothing like a charging problem. And finally the modules themselves, which is the expensive answer and the last one a good tech reaches for.
None of that is exotic work. It is methodical work, and it is why sending somebody with the right diagnostic tooling beats guessing at parts.
Why mobile and rail duty eats data links first
Here is the part that makes a railroad different from a pad in a parking lot.
A standby generator at a hospital sits on concrete and runs an exercise cycle once a week. Its harness lives a quiet life. A generator car runs down track, for hours, every day, in weather, with the engine loaded the entire time. Everything in that enclosure is being shaken continuously for the entire operating season.
Vibration does not usually break a connector outright. It works one loose over months. It wears insulation at a contact point until a conductor sees metal. It backs a pin out of a housing by a fraction until the connection is intermittent, which is the worst failure mode there is, because the unit works fine right up until it does not and then works fine again when the tech arrives.
Add temperature cycling and moisture and you get the same result faster.
So for anyone running gensets in mobile, rail, marine, or rental service, the maintenance emphasis is genuinely different from a standby fleet. Fluids and filters still matter. But the wear items that take these machines out of service are frequently electrical and mechanical connections rather than engine components. Harness inspection, connector integrity, mounting and isolator condition, and grounds are worth putting on the PM schedule as named line items, not as something a tech glances at while doing an oil change.
Two generators is not redundancy if you have not tested the second one
This is the part of the call worth the most to other operators, and it applies just as much to a hospital or a water plant as it does to a train.
The railroad did the right thing structurally. They carry a spare. One unit runs the load, the other is there for exactly the day the first one quits. On paper that is N+1 and it is the correct design for a business that cannot cancel a run.
But on the day it mattered, the primary was down and the spare had issues too, and the maintenance guy holding the phone was the one finding that out.
A backup that has not been started under load recently is not a backup. It is an assumption. The whole value of the redundant unit is that it works on the one day you need it, and the only way to know it works is to have run it, loaded, recently, and to have somebody’s name attached to the result.
For a two-unit setup in continuous service, a few things are worth being rigid about:
Alternate which unit carries the load. Running one machine for the whole season while the other sits is how you end up with one worn unit and one unproven one. Rotating them gives you two machines with known condition and roughly even hours.
Load test the spare, do not just start it. An engine that idles fine can still fail under real load, and HVAC compressors on a full train are real load.
Track faults on the standby unit the same as the running one. The second unit on this call had known issues that had not been resolved, because it was not the one holding the schedule up. That is completely understandable and it is exactly how a spare quietly stops being a spare.
Fix the spare before the season, not during it. Parts and technician availability are both better in the shoulder months than they are in the middle of a seven day a week operating schedule.
The windows between runs are the actual constraint
The caller flagged something most customers forget to mention until the tech is already on site: the train goes out two to three times a day, and the only access to the equipment is in the gaps between runs.
That is a scheduling constraint but it is really a diagnostic constraint, and it changes how a service call has to be run.
A tech working a normal standby job can take the machine out of service, work through the fault tree in order, and put it back together when he is done. A tech working between runs has a fixed window, cannot hold the equipment past it, and has to leave the car in a state where the train can move whether the repair is finished or not.
The practical answer is to front-load everything that does not require the window. Model and serial numbers, the exact fault codes on the display, what the mechanic already checked, what changed before it quit, and whether the fault is constant or intermittent. All of that can be gathered and sent before anybody drives out, and it determines which parts ride along in the truck. A comms fault where the tech shows up with connectors, pins, terminating resistors, and the right diagnostic interface is a different day from one where he shows up to look.
Two other things worth saying out loud when you call:
Tell the dispatcher the access window exists and how long it is. It changes who gets assigned and what they bring.
Tell them what the equipment is doing to the business right now. A machine that is inconvenient and a machine that is stopping revenue every hour get scheduled differently, and no service department can make that call on your behalf if you do not say it.
The stopgap on the back of the train
They finished the day’s runs on an older portable generator strapped to the back end.
That is the correct instinct and worth crediting. The passengers on that train had air conditioning and dinner, which is what they bought. Something running beats something correct on a Sunday afternoon.
It is also exactly the kind of arrangement that quietly becomes permanent, and it should not. A bridge unit chosen because it was the machine you happened to have is a machine nobody sized for the load, nobody load banked, and nobody is inspecting on a schedule. It buys a day. It does not buy a season, and the longer it stays on the back of the train the more likely it is to be there on a day when it also fails.
If you find yourself on a stopgap, put a date on it in writing. The repair timeline for the real unit is the date.
What operators should take from this
- Know that your genset has two controllers, not one. The engine ECM and the genset controller are separate, they talk over a data link, and that link is a real failure point with its own symptoms.
- A communication fault usually is not an engine fault. Gauges reading zero on a mechanically healthy engine points at the link, not the block. The repair is frequently a connector.
- Mobile and rail duty wears connections, not just engines. Put harness, connector, isolator, and ground inspection on the PM schedule as named items.
- Intermittent is worse than dead. A connection that works when the tech is standing there is the hardest and most expensive kind of fault to chase, which is an argument for capturing fault codes the moment they appear rather than after.
- A spare you have not loaded is not a spare. Rotate duty between units, load test the standby machine, and close out its faults even when it is not the one running.
- Say what your service window is when you call. It changes the assignment, the parts on the truck, and whether the trip accomplishes anything.
- Gather model, serial, and fault codes before the tech is dispatched. It is the cheapest thing you can do to shorten a repair.
- Put an end date on any stopgap. Temporary power that nobody sized and nobody tests is a second failure waiting for a worse day.
Frequently asked questions
What does a communication fault between the ECM and the generator controller mean? It means the engine control module and the generator set controller have lost the data link they use to talk to each other, typically a CAN bus running the J1939 protocol. The controller relies on the ECM for engine data such as speed, oil pressure, and coolant temperature, so when the link drops it may show zeroed gauges, a data link fault code, a failure to start, or a shutdown on a protection fault that did not actually occur. The engine itself is often completely healthy.
What causes a genset data link failure? Most commonly a connector problem, including corrosion, a backed-out pin, or a latch that has vibrated loose. Harness chafe or heat damage is next, followed by missing or failed CAN terminating resistors, then power and ground issues such as a weak battery or a corroded ground strap, which can produce communication errors that look unrelated to charging. Module failure is possible but is the least common cause and the last one a technician should conclude.
Why do mobile and rail generators fail differently from standby units? Because they run loaded for long periods while being vibrated continuously, and they see weather and temperature cycling that a pad-mounted standby set does not. That combination tends to wear electrical connections, harnesses, mounts, and isolators before it wears engine components, so maintenance programs for mobile fleets should treat connection integrity as a primary inspection item rather than an incidental one.
How often should a backup generator be tested if it is a spare unit? Frequently enough that its condition is a known fact rather than an assumption, and under actual load rather than at idle. For two-unit installations in continuous service, alternating which machine carries the load is usually better than designating one permanent primary, because it keeps hours even and confirms both machines work. Faults on the standby unit should be closed out on the same schedule as faults on the running unit.
What information should I have ready when I call for emergency generator service? The model and serial number of the unit, the exact fault codes showing on the controller, what the machine is doing and not doing, what was checked already and by whom, what changed immediately before the failure, whether the fault is constant or intermittent, and what access window the technician will have on site. That last one is routinely left out and it changes what a technician can accomplish on the trip.
Can a generator be serviced between operating runs? Often yes, but the window has to be known in advance because it determines what work is realistic. A technician who knows he has a fixed window will plan the sequence differently, bring the parts most likely to be needed, and leave the equipment in a movable state at the end of the window whether or not the repair is complete. Telling the service department about the constraint when you call is the difference between a productive trip and a diagnostic visit.
When the machine has to run tomorrow
Generator Source sells, services, rents, and buys industrial generators from 20kW to 4MW+, with branches in Brighton CO, Jacksonville FL, Pensacola FL, and Austin TX. Our field service runs on certified technicians and we take emergency calls, including on weekends, for equipment that cannot wait for Monday.
Talk to our service department or call the emergency line.