When the replacement roller develops the same flange wear at the same position, the second event gives the investigation a comparison the first repair lacked. The roller changed; the mounting relationship, neighbouring components and duty may have remained. The next repair needs to establish which of those retained conditions can explain the repeated contact, while checking that the two failures really are alike.
“It failed again” is still too broad to prove a common cause. The second event may involve the same installed position but a different damage mechanism, or the same component description on the opposite side under different exposure. A useful investigation begins by making the events comparable. Only then can the team decide which explanation survived the first repair and what evidence the next intervention must create.
Make the two events comparable before calling them repeats
Start with the physical object. Record the exact part identity available for each event, its side and position, the machine model and serial identity, and the date or exposure at installation and removal. “Roller,” “idler” or “track” is not enough. Two rollers in different positions can carry different loads and encounter different packing. A replacement that looks equivalent may have a different part number, revision or mounting detail.
Then describe the symptom without naming a cause. “Bearing failed” is often already an interpretation. More comparable observations would be leakage at a documented location, a roller that would not rotate, side wear on a flange, a damaged mounting surface, or measured play using an identified method. Photographs should include the whole position, neighbouring components and close details under similar orientation when possible.
Consider a fictional machine that develops lower-roller flange wear twice at the same right-side position. This is a comparison example, not an AFTparts service case. The useful record contains what was actually observed, including facts that fail to distinguish the events:
| Comparison | First event | Second event | What follows from it |
|---|---|---|---|
| Location and symptom | Right lower roller at the recorded position; inboard flange wear | Same position and same direction of flange wear | A recurring location and symptom are established |
| Time since installation | 180 engine hours | 160 engine hours | Both intervals use engine time; travel exposure is still unknown |
| Replacement identity | Removed roller's marking retained; original selection record missing | Replacement marking and supplier recorded; applicability not yet resolved | Changing supplier did not by itself test correct fit |
| First repair scope | Roller replaced | Same roller position now affected | The intervention changed the roller at that position |
| Mount and neighbouring parts | Retained; no dimensional inspection recorded | Still present; current inspection pending | Their condition is unknown, rather than confirmed acceptable |
| Work | Site assignment recorded; turning and travel not logged | Same site; task detail incomplete | Same site does not establish the same exposure |
The repeat at one position makes that local relationship worth investigating. The supplier change is also useful, but it is weaker evidence than it first appears: without an applicability check, both selections could share the same error. Meanwhile, the blank mounting inspection is a gap to close, not evidence that the mount caused the damage. A table that distinguishes those statements helps the team avoid spending a third repair on an untested assumption.
Exposure needs the same care as the object. Engine hours, travel hours, calendar time and production cycles are not interchangeable. If the first event was logged in engine hours and the second only by dates, the limitation belongs in the analysis. A recurrence after “about the same hours” is not established until the meter source, start and end points, and activity represented by those hours are comparable.
Change the investigation, not just the replacement
Once the event descriptions are aligned, sort potential contributors into three groups: what changed during the first repair, what remained in service, and what was never established.
Suppose a failed roller was replaced, but the neighbouring rollers, track path, mounting surface, track tension and packing conditions were not recorded. The new roller changes one component. It does not test whether a damaged mount is imposing misalignment, whether a neighbour is affecting the path, or whether material is repeatedly preventing rotation. Those are not automatically the causes; they are explanations the first intervention did not eliminate.
Turn each suspected cause into a test
The investigation should now ask questions that can fail. “Check alignment” is too vague. A better question is whether an inspection under the applicable machine procedure shows the component and track path within the specified relationship. “Operator abuse” is not a useful hypothesis. A testable version identifies the particular duty or event expected to create the observed direction and location of damage, then compares that expectation with operating records and the opposite side of the machine.
Each proposed explanation should connect three elements:
- Mechanism: the physical way the condition could create this damage.
- Expected evidence: what should also be visible or measurable if the mechanism is active.
- Contradicting evidence: a result that would make the explanation less likely.
In the example, start with the hypothesis that an incorrect replacement placed the running surfaces in the wrong relationship. The expected evidence is a discrepancy between the installed identity or geometry and the applicable machine information. A documented match weakens that explanation. It does not prove the whole undercarriage is sound, but it closes a specific uncertainty that a supplier change left open.
The mounting hypothesis asks a different question: does the retained mounting relationship meet the machine's specified condition? An applicable inspection showing a discrepancy would give the hypothesis physical support. A missing inspection record merely explains why it remained untested. For a duty-related hypothesis, look for a recorded manoeuvre or load consistent with the side and direction of wear; do not deliberately reproduce damaging operation to see whether the symptom returns.
The repair history must also be examined as evidence, not treated as proof of correct work. Was the replacement identity independently confirmed? Were mating surfaces inspected? Was the installation condition documented? Were the relevant settings checked under the exact machine information? A signed work order can establish that a task was recorded. It cannot establish the unrecorded condition of every neighbouring component.
Recurring-failure guidance from Limble and Pros at Work emphasizes moving beyond symptom replacement toward causes that keep producing the event. Shop and Supply's heavy-equipment failure-analysis discussion similarly frames the failed part as evidence within a broader system. Those sources support the investigation logic, but they do not diagnose this machine. The applicable manufacturer procedures and physical findings remain controlling.
Define evidence that would show the corrective action worked
A corrective action is incomplete if success means only that the machine returned to service. The replacement removes the failed item; the corrective action is meant to change the condition believed to have produced the failure. Those two actions can be the same only when the component itself is supported as the cause and the surrounding system has been excluded appropriately.
Write the expected result before enough time passes to forget the original hypothesis. If the action corrected a mounting relationship, the record should identify the applicable check, its result and who completed it. If packing was addressed, define what will be inspected and under which comparable conditions. If the suspected driver was a particular duty cycle, record exposure to that duty rather than relying only on total engine time.
The follow-up needs to observe the proposed correction itself as well as the damaged part. A newly fitted roller will initially present an unworn flange regardless of whether the mounting relationship was corrected. A photograph of that clean flange is therefore a baseline, not evidence that the cause has gone. The mounting check and subsequent contact pattern answer the effectiveness question.
For the fictional flange-wear recurrence, the technician would record the verified repair condition and the same flange view at handover. The operator would report a return of the observed contact or tracking symptom through the normal maintenance process. The planner would arrange a comparison at an exposure selected under the applicable service guidance, taking account of the previous 160–180 engine-hour windows and the seriousness of the condition. Those historical numbers frame this example; they are not a recommended inspection interval. A review before comparable exposure, and a further review after comparable work, answer different questions about whether the correction is holding.
The follow-up should be tied to meaningful exposure and the expected development of the symptom. That does not mean inventing a universal number of hours. A slow dimensional trend and an intermittent derailment require different evidence. The applicable service information, severity of the condition and previous exposure window should guide the interval.
No recurrence during one short observation window does not prove the cause was removed. The machine may not have encountered the same ground, travel load or operating pattern. Conversely, an early repeat after a documented change can strongly challenge the hypothesis if the two exposure windows are comparable. This is why conditions and meter definitions belong in the effectiveness check.
If the investigation points to a selection error, compare the removed and replacement track roller identities with the machine and mounting information. If it points to a retained mechanical condition, correcting that condition belongs in the repair scope alongside the damaged roller. The two findings lead to different work even though both may involve another replacement part.
At the follow-up, compare the same location, symptom and exposure definition with the two earlier events. A sound result includes both the changed physical condition and its behaviour under relevant work. If the symptom recurs, retain that result too: it tells the next investigation that the documented intervention did not remove everything producing the pattern.