Diagnosing chronic cracking issues on a 3-tonne excavator boom is challenging. CMP combined high-rate field instrumentation with a 3D-scan-driven FEA model to move the conversation from the typical “where is it cracking” to “what factors drive instantaneous damage” and used that insight to engineer a repair predicted to lift fatigue life of the critical weld by more than 7×.
Background
- Why was the existing 50 mm OEM repair not lasting?
- What repair geometry would actually fix it
- How would the team know the fix would achieve the predicted life?
Challenge
- The loading was heavily duty-cycle dependent, with overhand vs underhand digging, single vs double benching, and operator technique all changing the stress spectrum.
- Discrete "lateral impact" events (bucket engaging the bench mid-slew) generated significant stress spikes that could dominate any time-history plot but might or might not have been the actual driver of fatigue damage.
- The cracking initiated near a transverse butt weld at the boundary of the previous OEM repair (which was a known poor weld detail under weld fatigue design codes), but the available evidence of cracking after the fact gave no quantitative basis for choosing a new repair geometry.
- Conventional doubler-plate repairs commonly applied to R996 boom feet were suspected of merely shifting the crack front into the existing parent plate, which already contained weld defects from prior repairs.
Solution
The excavator was instrumented with a synchronised data acquisition system capturing the structural response and the machine kinematics simultaneously:
- Eight strain gauges on the boom foot (four on the upper surface, four on the lower surface) were positioned to capture corner stresses where cracking initiated.
- Hydraulic pressure transducers on the boom-lift, stick and bucket-tilt cylinders.
- Rotary encoders on the boom and stick to recover joint angles.
- A six-axis IMU on the machine providing three-axis linear acceleration (Aₓ, Aᵧ, A_z) and three-axis rotational rate (Rₓ, Rᵧ, R_z).
2. Time-synchronised correlation of stress with machine state
This was where the investigation departed from a conventional strain-gauging exercise. Because every stress sample carried a matched record of cylinder pressures, joint angles and inertial rates, CMP Engineers could correlate the stresses at any point in time with knowledge of what the machine was doing; what angle the boom, stick and bucket were at, what force was being applied by the hydraulic cylinders, and when dynamic impact events occurred.
The dominant “high-stress” events (transient excursions to ±200 MPa at a single gauge) were initially suspected to be the primary fatigue driver. Cross-referencing the IMU and cylinder data showed, however:
- No coincident spike in cylinder pressures >> The event was not due to operators overfilling the bucket
- No spike in translational acceleration >> The event was not a drop or pothole impact.
- A significant spike in rotational rate about the vertical axis >> The event was a swing-deceleration transient.
This isolated the failure-mode-relevant event as the bucket engaging the bench before the swing motion had fully halted and ruled out several other plausible explanations based on hard data rather than assumption.
3. Key innovation: Spatial stress mapping
Conventional strain-gauging reports stresses at discrete location and time, and not much more. CMP went a step further. The boom and stick joint-angle data (from the rotary encoders) and the machine geometry were used to recover the stick-tip position in the machine frame at every sample. Each strain-gauge reading was then plotted at the stick-tip location at which it occurred, producing a stress heat-map across the entire operating envelope of the machine for each shift recorded.
The results, with one example shown in Figure 3, made the failure mechanism legible at a glance. The hottest peak-stress regions at SG2 (lower boom-foot corner) lie at the outer-reach edge of the dig envelope, close to ground level. This confirmed, using the observed data, that the damaging events occurred when the operator was digging with the boom and stick extended out near their reach limits.
This visualisation did three things that no conventional time-history plot could:
- It collapsed tens of millions of samples into a single, decision-ready picture for site engineering.
- It localised the damaging events in machine-frame coordinates that operations and training teams understood intuitively (“don’t dig out there with the bucket at ground level“) rather than in cryptic time-domain line plots.
- It provided a basis for future controls work: the same fusion of strain, IMU and joint-angle data was the substrate on which operator-coaching feedback, machine envelope limits, or onboard structural health monitoring could later be built.
4. Rainflow fatigue analysis
The full time-history stress data at each gauge was processed via rainflow cycle counting to produce a stress-range/cycle-count spectrum. This was then converted to fatigue damage using the BS 7608 S-N curves for the relevant welded-joint class. The headline conclusion was, perhaps counter-intuitively, that the rare ±200 MPa lateral-impact events were not the dominant fatigue driver. Instead, the routine dig-swing-dump-return cycle at ~100 MPa stress range, by virtue of its frequency, was the driving factor in fatigue. Predicted mean times to crack initiation at each gauge were found to correlate well with field history.
5. 3D-scan-based FEA correlation
A 3D scan of the boom was used as the geometry baseline for a Finite Element Analysis (FEA) model that included simplified representations of the stick, bucket and cylinders, so that inertial loadings and bucket reaction loads could be applied realistically. Multiple load cases, including start-of-dig, dump, and lateral-impact, were evaluated, with the boundary conditions taken directly from the matched field data.
Reasonable correlation was obtained between the FEA-predicted stress ranges and the strain-gauge measurements, giving confidence that the model could be used predictively to compare repair options.
Outcomes
Root cause identified on evidence, not gut feel
The cracking was attributed to the transverse butt weld at the boundary of the previous OEM 50 mm thickened-plate repair, with the dominant fatigue driver being the high-frequency ~100 MPa normal dig cycle and the worst stresses occurring in the outer-reach-at-ground zone of the operating envelope.
Repair options assessed, and one explicitly ruled out
Doubler plates (the typical site-level fix on R996 boom feet) were assessed and recommended against. The existing parent plate retained weld defects from prior repairs, and adding a doubler simply shifts the initiation site without removing the defect-bearing material.
Repair designed with a quantified life prediction
CMP proposed an 80 mm thickened-plate repair, with the plate footprint extended at both ends so that the transverse welds sit well clear of the high-stress zones identified in the FEA.
FEA of the proposed repair predicts:
Why this matters
Beyond the immediate repair, the work produced two outputs of broader value to the operation:
- A repair package with a defensible engineering basis. The 7.25× predicted-life improvement at the critical weld was grounded in matched in-service data and a correlated FEA model, not in OEM precedent or rule-of-thumb scaling.
- A reusable analysis framework. The fusion of strain, IMU, cylinder pressure and joint-angle data (and in particular the spatial stress-mapping technique) was directly transferable to any other mining shovel, dragline or large mobile structure where fatigue life was operator- and duty-cycle-driven. The same data pipeline could underpin onboard structural health monitoring, operator coaching, or duty-cycle benchmarking for fleet-wide decisions.
Capabilities demonstrated
