Eddy-current proximity probes and accelerometers answer different physics questions on the same machine: a proximity probe measures shaft-relative displacement (typically 0-500 mil / 0-12.7 mm full scale, 200 mV/mil or 7.87 V/mm sensitivity), while an accelerometer measures absolute housing acceleration (IEPE, g peak / g RMS, 0.5-10 kHz flat band) [S1][S3][S5].
Both technologies can drive early fault detection on rolling-element bearings, but only when the signal chain — cable length, target material, gap voltage, and ICP excitation current — is specified to the rotating assembly they sit on [S1][S3].
Measurement Frame: Shaft-Relative vs Housing-Absolute
A proximity probe is a non-contact eddy-current transducer: the controller drives the tip coil, the rotor surface sets the gap, and the output is a calibrated DC voltage proportional to the air gap (linearity is specified over a 0.5-2.5 mm range depending on the 5 mm, 8 mm, 11 mm or 25 mm system) [S3][S5].
An accelerometer is a contact piezoelectric device bolted to the bearing housing; its output represents the casing's inertial response to the bearing, not the shaft itself, so casing resonances above ~1-3 kHz contaminate the bearing-fault band unless the mounting is rigid and the surface is flat [S1]. The practical consequence is that the same bearing defect will show up in displacement on the shaft and as a high-frequency demodulated envelope on the housing, with the two signals being mathematically related by double integration only below the casing-resonance cut-off [S1].
Selection Criteria: When to Specify Each
Specify a proximity probe when the API 670 / ISO 10816 protective-monitoring requirement is shaft-relative vibration, axial position, differential expansion, or case expansion on steam turbines, hydro turbines, and large process-gas compressors [S5]. A 4-20 mA proximity transducer variant is used for slow-moving trip channels (radial vibration, axial thrust, thermal growth), while a dynamic probe feeds the rack monitor for waveform and 1×/2× amplitude alarms [S5].
Specify an accelerometer when the asset is mid-speed, the bearing housing is accessible, the fault of interest is bearing-race spalling or gear-mesh wear, and the data will go through envelope (HFRT) demodulation or to a portable data collector for periodic route-based surveys [S1]. For high-frequency bearing and gear diagnostics, even proximity-based displacement data has been shown to be effective — but only after EEMD decomposition and an impulsiveness-indicator step that isolates the fault-sensitive IMF [S1].
Do not specify either on its own for a critical train: the protective channel (shaft-relative, 4-20 mA proximity) and the diagnostic channel (housing acceleration, IEPE) are independent channels, and protective-only systems miss the very slow-growing bearing race defects that diagnostic-only systems miss rotor bow and rubs [S1][S3].
Comparison: Decision Matrix Across Four Criteria
On the four criteria that drive instrument-engineer selection, the contrast is sharp. <strong>Reference frame:</strong> proximity = shaft-relative displacement; accelerometer = housing-absolute acceleration [S1][S3]. <strong>Frequency band:</strong> proximity = 0-1 kHz useful dynamic (DC to 1×-10× running speed); accelerometer = 0.5 Hz-10 kHz usable, with bearing/gear energy typically read above 5 kHz after envelope demodulation [S1]. <strong>Installation surface:</strong> proximity needs a machined, eddy-current-friendly target (typically 4140 / 17-4PH shaft journal with Ra ≤0.8 µm, calibrated 1.0-1.5 mm gap, 80-100 mV/mil mid-scale); accelerometer needs a flat, rigid mounting pad (≤5° mounting error for cross-axis rejection) and a fixture natural frequency above 10× the analysis band [S3]. <strong>Failure cost:</strong> an undetected proximity-probe impedance error produces false trips or missed trips on the protective channel — a "catastrophic" outcome per the OEM's own field literature — while a degraded accelerometer sensitivity produces a quiet, slow under-reporting of bearing wear that only shows up in post-failure teardown [S3].
Calibration and Field Validation
Portable vibration calibrators now expose dedicated proximity-probe adaptors so that a field tech can validate mid-scale sensitivity (mV/µm or mV/mil) and linearity in the actual installed gap, not just on the bench [S3]. The OEM guidance is that "gap voltage alone is incapable of detecting" wiring and impedance faults, so the recommended workflow is: bench calibration → install at the calibrated gap → re-verify sensitivity in-situ with the portable calibrator → document the reading into the CMMS record [S3].
For an ICP accelerometer, the equivalent workflow is a 100 Hz @ 1 g reference check on a portable shaker, plus a capacitance-mounted resonance sweep whenever the mounting stud is changed [S3]. The Modal Shop, CTC, Brüel & Kjær, and similar suppliers publish the adaptor list (probe diameter, tip thread, target material fixture) so that a single calibrator covers the full fleet of probe sizes on site [S3][S5].
Use-Case Recommendations by Machine Class
For steam turbines, hydro turbines, and large API 617 process compressors, the protective channel is non-negotiably a shaft-relative proximity probe (radial + axial, with 4-20 mA trip), and the diagnostic channel is an accelerometer pair on each bearing housing, fed to a rack monitor that records both waveforms [S5]. The cost of a single 8 mm or 11 mm probe assembly plus extension cable is small relative to the trip-system the probe feeds; refurbished spares such as the TEMP-Pro 10-0845-001 housing assembly appear on the secondary market at roughly US$30-35 each, but field engineers typically require a documented bench calibration certificate before installing a used assembly [S2][S3].
For mid-speed centrifugal pumps, small fans, and motor-pump skids, a single high-quality IEPE accelerometer per bearing is the usual minimum: shaft-relative measurement is impractical because the impeller end is sealed, and a non-contact probe cannot survive the oil-mist environment [S1]. For rolling-element bearing fault detection on any machine, a 2023 study in the Journal of Vibration Engineering & Technologies confirmed that proximity-based displacement data, after EEMD decomposition, can match accelerometer-based envelope analysis on damaged-bearing test rigs — so the diagnostic channel choice is increasingly a question of which signal is already in the rack and which bandwidth the analyst is trained on, not a question of physics [S1].
Limits, Failure Modes, and Standards Discipline
Proximity probes are sensitive to shaft material change (4140 vs 17-4PH vs 410 stainless calibrations are not interchangeable), to probe-cable capacitance above the OEM limit (cable + extension cable > 100 m typically degrades the 0-10 kHz driver), and to target surface damage that the gap-voltage reading will not flag on its own [S3][S5]. Accelerometers are sensitive to mounting-stiffness loss (a hand-held probe halves the high-frequency amplitude by 200-1000 Hz), to ICP current limit on long cables, and to thermal transient on the piezoelectric crystal above the published upper temperature [S1].
Protective-monitoring conformance on turbomachinery typically references the API 670 vibration-and-position-monitoring standard, while diagnostic conformance on rolling-element bearings is dominated by ISO 15243 (failure modes) and ISO 10816 (machine vibration evaluation zones). Diagnostic algorithms for both sensor types are published alongside, not inside, the sensor standards, which is why displacement-based bearing fault methods like EEMD still require a peer-reviewed validation per asset class before they are accepted as a replacement for envelope-demodulated accelerometer channels [S1].
Where This Leaves the Spec Engineer
Specifying neither is rarely correct; specifying one instead of the other is a budget cut, not an engineering choice. The cost gap between a 4-20 mA proximity probe channel and an IEPE accelerometer channel has narrowed to the point where the dominant differentiator is the analyst's training, the rack monitor's channel count, and the field tech's ability to re-validate the probe at its installed gap. Three near-term signals to watch: more portable calibrators advertising native probe-adaptor sets [S3], more EEMD-style bearing-fault algorithms validated on shaft-relative displacement rather than only on housing acceleration [S1], and more proximity probe selection guides that explicitly model the 4-20 mA protective channel alongside the dynamic waveform channel. For context on the surrounding cost stack, the condition monitoring system price and cost guide and the vibration analyzer price 2026 breakdowns both treat the proximity + accelerometer pairing as the default protective-plus-diagnostic architecture for turbomachinery, not as alternatives.
For component-level specifications, see proximity sensor.