Acoustic emission testing has been used in structural integrity assessment — pressure vessels, weld inspection, aerospace components — since the 1960s. Its application to rotating machinery condition monitoring is more recent, and for most reliability engineers in heavy industry, it remains less familiar territory than vibration analysis or oil sampling. That's worth correcting, because AE is genuinely one of the earlier-detecting technologies available for bearing and gear defects, and the barrier to understanding it is lower than it might appear.
This article covers the physics, the practical deployment realities, and what AE can and cannot detect. No lab equipment required — this is written for the reliability engineer deciding whether to add AE to their condition monitoring toolkit, not for the NDT specialist.
What Acoustic Emission Actually Is
When material undergoes rapid localized stress release — a crack propagating, a contact fatigue spall initiating, metal-to-metal asperity contact under inadequate lubrication — it releases energy in the form of elastic stress waves. These waves propagate through the material at the speed of sound and can be detected at the component surface by a sensor sensitive to the appropriate frequency range.
The critical distinction from vibration analysis is the frequency range. Standard vibration accelerometers operate from roughly 10 Hz to 10–20 kHz. Acoustic emission sensors operate from 100 kHz to 1 MHz or higher. This difference in frequency range is not incidental — it determines what physical phenomena each technology can detect.
Low-frequency vibration arises from the dynamic response of the entire machine structure — imbalance, misalignment, looseness, and (at later stages) bearing defect impacts that have sufficient energy to excite structural resonances measurably. High-frequency acoustic emission arises from localized stress events at the micro-structural level: crack tip propagation, contact asperity failure, micro-plastic deformation. These are different physical phenomena, and they occur at different stages of the degradation timeline. AE events occur first.
The Defect Detection Timeline
For a rolling element bearing developing a raceway fatigue spall, the AE detection window looks roughly like this:
- Subsurface crack nucleation: The crack has initiated below the surface. Stress waves are being emitted at low amplitude. AE sensors at close range may begin detecting a change in emission rate. Vibration: no change. Oil: no change. This is Stage 0 — some AE systems can begin flagging this, most are not sensitive enough at typical installation distances.
- Crack growth toward surface: The subsurface crack grows under repeated contact stress. AE emission rate increases and begins showing bearing-defect-frequency-related repetition in the time-domain signal. Sensitive AE monitoring will detect this as an anomaly. Vibration: still at or near baseline. Oil particles: beginning to show trace ferrous content increase.
- Surface breakthrough and early spall: The crack breaks the raceway surface. Each rolling element passage over the spall generates a burst-type AE event. The AE signal becomes clearly periodic at the bearing defect frequency (BPFO for outer race, BPFI for inner race). High-frequency envelope demodulation on vibration may now begin showing weak sidebands. This is where well-implemented vibration analysis starts to see something. AE is already well ahead.
- Spall propagation: Both AE amplitude and vibration increase. The condition is now detectable by multiple methods. Planning window: weeks to a few months depending on load and speed.
- Accelerated damage: Macro-spalling, particle contamination in the lubricant, rapid vibration amplitude increase. Emergency intervention window: days to weeks.
For a typical industrial bearing — say, the DE bearing on a 75 kW centrifugal pump running at 1480 RPM — the difference in detection horizon between AE and vibration tends to be in the range of 4–10 weeks in our experience with this equipment class. That's not a guarantee; it depends heavily on the specific failure mode, load severity, and sensor installation quality. But it represents a consistent pattern.
AE Parameters You Need to Understand
Hit Rate
An AE "hit" is a discrete detected signal that exceeds a threshold. Hit rate (hits per second or hits per revolution) is a basic indicator of AE activity level. Rising hit rate at consistent operating conditions is a degradation indicator. The limitation: hit rate alone can't distinguish bearing defects from other AE sources in the machine (cavitation, seal contact, gear meshing).
Amplitude Distribution
AE amplitudes are typically expressed in dB relative to a reference voltage (1 microvolt at the sensor terminals). Healthy bearings show an amplitude distribution skewed toward lower amplitudes. As defects develop, the distribution shifts right — more high-amplitude events. The b-value analysis (borrowed from seismology) tracks the slope of this amplitude distribution and is a useful metric for tracking degradation rate.
RMS and Energy
RMS of the AE waveform is analogous to vibration RMS — it integrates the overall activity level into a single trending value. AE energy (proportional to RMS squared integrated over time) is more sensitive to large individual events. Both are useful for trending; energy is often more sensitive for early detection.
Time-Domain Waveform and Envelope
For bearing defect frequency identification, you need the time-domain AE waveform sampled at sufficient rate (typically 1 MHz or above for proper reconstruction of burst events). Envelope analysis on the high-frequency AE signal — analogous to envelope demodulation in vibration analysis — extracts the repetition rate of burst events and allows identification of bearing defect frequencies in the AE signal. This is where AE analysis starts to look like familiar vibration analysis territory.
Sensor Installation: Where Most AE Programs Fail
AE technology works in the lab. It frequently disappoints in the field. The most common reason is poor sensor installation. At frequencies above 100 kHz, the signal attenuation through interfaces — paint, rust, uneven surfaces, air gaps — is severe. A sensor mounted on a painted surface with inadequate coupling can have 20–30 dB less signal than the same sensor properly coupled to clean bare metal. That 20–30 dB loss corresponds to moving from early-stage defect detection capability to detecting nothing until the damage is severe.
Practical installation requirements:
- Contact surface ground to bare metal, Ra < 1.6 µm if possible
- Couplant appropriate for temperature: silicone grease below 80°C, high-temperature grease above that
- Magnetic or permanently bonded mounting preferred over clamps for continuous monitoring
- Sensor as close to the bearing race as possible — each additional centimeter of signal path through structural metal attenuates the signal
- Cables protected from mechanical contact and electromagnetic interference sources (motor windings, VFDs)
We're not saying AE requires exotic installation procedures — it doesn't. But it does require more attention to surface preparation than bolting an accelerometer to a painted housing does. Shortcuts here directly compromise detection capability.
What AE Cannot Do
AE is not a universal replacement for vibration analysis. There are fault modes and machine types where AE provides no advantage, and some where it's actually disadvantaged.
Unbalance and misalignment: These are sub-synchronous and synchronous fault signatures that manifest in the vibration frequency domain. AE does not detect them. If you're looking to identify that a machine is misaligned, vibration spectrum analysis is the right tool.
Structural looseness: Foot looseness, baseplate issues, and structural resonance problems are vibration domain problems. AE will detect the metal-to-metal contact events associated with looseness, but it won't tell you where in the structure the problem lies.
High-noise environments: AE sensors are sensitive, and in environments with continuous high-frequency structural noise — near-field impact from high-pressure fluid jets, directly adjacent pneumatic systems, high-speed gear meshing with known tooth damage — signal-to-noise ratios can make bearing defect detection unreliable. Machine-learning-based signal classification helps here but requires training data from known conditions on the specific machine.
Very large machines at distance: AE signal attenuation in metal is roughly 2–4 dB per centimeter at 150 kHz for typical steel structures. On a large machine — a multi-stage gearbox, a large compressor — getting sensors close enough to all critical bearing positions to maintain detection sensitivity may require multiple sensors and careful attenuation mapping.
Where AE Fits in a Multi-Technology Program
The most defensible position for AE in a reliability program is as the early-warning layer for rotating equipment bearing condition, particularly for assets where the cost of missed detection is high and the P-F interval is short relative to your maintenance planning cycle. It works alongside vibration analysis, not instead of it.
For a plant with 200 rotating assets, you probably don't AE-monitor all of them. You identify the 20–30 assets where: (a) bearing failure consequence is highest (production loss, safety), (b) bearing failure has occurred with insufficient warning time under vibration-only monitoring, or (c) operating speed is low enough that vibration analysis is genuinely unreliable. Those are your AE candidates. The rest stay on your existing vibration and lubrication program.
The question isn't whether AE is better than vibration. It's whether the additional detection horizon AE provides, on the specific assets where vibration monitoring has historically given you inadequate warning time, justifies the additional instrumentation and analysis investment. For the right asset class, the answer is consistently yes.