01
Mechanical change
Mechanical changes affect load
Bearing wear, imbalance, cavitation, misalignment, and process changes can alter torque, speed, or load. The driven machine starts asking the motor to work differently.
Electrical Signature Analysis
Read asset condition from the motor's electrical signal.
Electrical Signature Analysis measures current and voltage at the motor control cabinet, then analyses the signal for fault patterns linked to the motor, drivetrain, and driven load. SAM4 uses ESA to monitor hard-to-reach motor-driven assets without installing sensors on the machine.
Why mechanical faults can appear in electrical data
02
Motor response
The motor responds electrically
The motor converts electrical energy into rotation through the magnetic field between stator and rotor. When load or magnetic balance changes, current and voltage can change in small but repeatable ways.
03
Cabinet measurement
SAM4 measures at the cabinet
SAM4 captures current and voltage at the motor control cabinet and analyses the signal in the frequency domain. No sensor is needed on the asset.
The point. Faults that affect torque, speed, load, or electromagnetic balance can leave repeatable patterns in current and voltage. SAM4 reads those patterns at the MCC and turns them into fault evidence.
How fault mechanisms become electrical evidence
An induction motor links the electrical supply to the mechanical system. When the motor, drivetrain, or driven load changes behaviour, the motor can draw current differently. ESA reads those changes at the motor control cabinet.
Causal chain
01
Mechanical or process change
Bearing wear, cavitation, misalignment, blockage, belt slip
02
Motor response
Torque, speed, load, electromagnetic balance
03
Electrical signature
Current and voltage waveform changes
04
Spectral analysis
Frequency components, harmonics, sidebands
05
SAM4 finding
Fault evidence, severity, recommended action
The motor is the coupling point
Inside the motor, the rotor and stator are separated by a narrow air gap. In healthy operation, magnetic flux across that gap follows a stable pattern for a given speed and load.
Motor-internal faults can disturb that electromagnetic balance. A worn bearing may shift rotor position. Rotor bar damage can alter the magnetic field. Winding degradation can affect phase behaviour.
Downstream faults are also detectable, but through a different path. Cavitation, blockage, belt slip, gearbox wear, or misalignment can change torque demand, speed stability, or load. The motor responds electrically to those changes.
From fault to electrical signature
Fault-related changes can create small, repeatable variations in current and voltage. In the frequency domain, those variations appear as components, harmonics, or sidebands around expected electrical and mechanical frequencies.
The signature depends on the asset, load, speed, drive configuration, signal path, and fault severity. That is why SAM4 combines physics-based fault models with asset-specific baselines and operating context. The two-track analysis section below describes how SAM4 separates fault patterns from operating context.
How SAM4 reads the signal
Cabinet-installed measurement hardware captures current and voltage from the motor supply. The waveforms are sampled at high frequency and transformed into spectral features using methods such as Fast Fourier Transform and wavelet analysis.
The takeaway: the motor couples electrical and mechanical behaviour. Faults that change torque, speed, load, or electromagnetic balance can leave repeatable signatures in current and voltage. SAM4 reads those signatures at the cabinet and validates them against asset context. Detection strength varies by asset type, operating regime, load stability, and fault mechanism.
How electrical, drivetrain, and load-path faults change the signal
A motor-driven system links the electrical supply to the driven load. Electrical energy flows forward through the drive, motor, coupling, and machine. Fault evidence can appear in the opposite direction: mechanical and process changes alter torque, speed, load, or electromagnetic balance, and those changes can show up in current and voltage at the motor control cabinet.
01
Fault develops
Faults can originate in the motor, coupling, gearbox, belt, pump, fan, compressor, or process.
02
Motor response changes
The fault changes torque demand, speed stability, load, drag, slip, or electromagnetic balance.
03
Electrical signature changes
Those changes can create repeatable components, harmonics, or sidebands in the current and voltage spectrum.
04
SAM4 measures at the MCC
SAM4 captures high-frequency current and voltage at the motor control cabinet and analyses the signal for fault evidence. No sensor is installed on the asset.
Power flows forward. Fault evidence can appear backward through the motor's electrical behaviour. SAM4 reads that evidence at the MCC.
Detection is asset-specific.
Evidence strength varies by asset type, fault mechanism, operating regime, load stability, drive configuration, and signal path. SAM4 reports performance by asset type instead of relying on one blended number across all machines.
See asset-specific recall figures →How SAM4 turns spectra into fault evidence
SAM4 captures high-frequency current and voltage waveforms at the motor control cabinet. Those waveforms are transformed into spectral features, where many electrical, mechanical, and process-related fault patterns become easier to separate. SAM4 then runs two analysis tracks in parallel: one looks for known fault signatures grounded in physics, and the other compares each asset against its own healthy baseline.
Current + voltage waveforms
↓
Spectral features
↓
Physics-based track
Known fault signatures
- Expected components and harmonics
- Sidebands and modulation patterns
- Linkable to a named fault mechanism
Baseline track
Asset-specific normal behaviour
- Drift from each asset’s own baseline
- Novelty across operating context
- Useful for unusual or combined patterns
↓
Reliability engineer validation
↓
Validated finding
Example fingerprints, grouped by detection path
A non-exhaustive sample of patterns SAM4 looks for. Specific frequencies, harmonics, and indicators depend on asset type, drive configuration, and operating regime.
Direct electrical
Stator winding faults
Can increase negative-sequence current and selected harmonic indicators.
Broken rotor bars
Can produce sidebands around the supply frequency related to motor slip.
Motor / air-gap
Air-gap eccentricity
Can produce rotor-slot and running-speed-related components.
Bearing defects
Can create modulation patterns associated with BPFO, BPFI, BSF, and FTF.
Load-path / process
Misalignment
Can create rotational-frequency components and sidebands, often involving 1× and 2× running speed.
Cavitation
Can raise broadband energy and create unstable load-related patterns.
Why voltage matters too
Current shows how the motor responds to load, torque, speed, and electromagnetic changes. Voltage helps separate those effects from supply-side disturbances such as voltage imbalance, harmonic distortion, and grid-related variation. Measured together, they let SAM4 distinguish a developing asset fault from a power quality issue.
Detection confidence is asset-specific.
Signal strength depends on asset type, load stability, operating regime, drive configuration, and fault mechanism. SAM4 reports performance by asset class rather than relying on one blended metric.
See asset-specific recall figures →What SAM4 can detect from current and voltage
SAM4 has detected the fault classes below in operating assets. Detection confidence depends on the asset type, drive configuration, duty cycle, load stability, signal quality, and whether the fault creates a measurable change in current, voltage, torque, speed, or load.
Use this table as a detection-fit guide, not a blanket coverage guarantee. The strongest candidates are assets where the fault changes the motor's electrical signature through load, torque, speed, hydraulic behaviour, or electromagnetic balance.
Fleet-level evidence means SAM4 has a substantial scored cohort for this detection class, so we report a pooled performance metric across the audit window.
Case-level evidence means SAM4 has detected this fault class in the field, but we report the evidence per asset and case rather than as a pooled headline metric.
Detection performance depends on asset type, operating regime, drive configuration, signal quality, and fault mode.
| Detection class | Signal path | Best-fit assets | Evidence basis | Boundary |
|---|---|---|---|---|
| Clogging / blockage | Process / load-path | Pumps, submerged sewage pumps | Fleet-level evidence | Strongest when blockage changes load, hydraulic behaviour, or duty pattern. |
| Process-induced load deviation | Process / load-path | Pumps, fans, mixers, conveyors | Fleet-level evidence | Detects load-pattern changes, not every process deviation. |
| Mechanical unbalance | Electromechanical | Pumps, fans, conveyors | Fleet-level evidence | Strongest when unbalance affects torque, speed, current, or voltage signature. |
| Voltage imbalance | Direct electrical | Motor-driven assets | Fleet-level evidence | Applies where phase imbalance is visible in the measured supply voltage. |
| Belt degradation | Transmission / load-path | Belt-driven systems | Fleet-level evidence | Strong where wear, slip, or tension loss changes speed or load behaviour. |
| Belt misalignment / tracking issue | Transmission / load-path | Belt-driven systems | Fleet-level evidence | Visible when misalignment affects load, friction, speed, or belt behaviour. |
| Fouling or contamination causing load change | Process / load-path | Fans, conveyors, compressors | Fleet-level evidence | Indirect; detects the load effect, not contamination itself. |
| Bearing degradation indicators | Electromechanical | Pumps, conveyors, mixers, blowers | Case-level evidence | Detection depends on asset type, fault location, speed, load, and signal path. |
| Shaft or coupling misalignment | Electromechanical | Pumps, fans | Case-level evidence | Strongest when misalignment changes torque, speed, load, or current signature. |
| Cavitation | Process / hydraulic load | Pumps | Case-level evidence | Visible when cavitation changes hydraulic load or operating pattern. |
| Air lock | Process / hydraulic load | Submerged sewage pumps | Case-level evidence | Strongest when air lock creates a distinct load or hydraulic signature. |
| Coupling-related load anomaly | Transmission / load-path | Pumps, conveyors | Case-level evidence | Visible when coupling mass or alignment effects appear as torque, speed, or load changes. |
| Gearbox degradation or gear-mesh anomaly | Transmission / load-path | Conveyors, fans, mixers, pumps | Case-level evidence | Detection depends on gearing, load path, fault progression, and signal strength. |
| Impeller degradation | Process / load-path | Pumps | Case-level evidence | Indirect; strongest when degradation creates sustained hydraulic or efficiency changes. |
| Seal-related load anomaly | Process / load-path | Pumps | Case-level evidence | Not direct leak detection; visible only when seal issues create measurable load, hydraulic, or electrical effects. |
| Pulley degradation | Transmission / load-path | Belt-driven systems | Case-level evidence | Strongest when pulley wear affects belt speed, slip, tension, or load. |
| Soft foot indicators | Electromechanical | Fans | Case-level evidence | Detection depends on whether mounting distortion creates a measurable electrical or load effect. |
| Stator winding short indicators | Direct electrical | Motor-driven assets | Case-level evidence | Detection depends on severity, signal quality, and electrical configuration. |
| Overloading | Process / load-path | Compressors, conveyors | Case-level evidence | Detects sustained or abnormal load patterns, not every short transient. |
What SAM4 does not detect directly
SAM4 does not directly measure flow, pressure, vibration, temperature, product quality, or process chemistry. It detects faults when they create measurable changes in current, voltage, torque, speed, load, or electrical balance.
Some failure modes remain better covered by vibration, process instrumentation, OEM monitoring, inspection, oil analysis, or offline electrical testing.
Recall = FD / (FD + FM). False-alert rate = FC / (FD + FC). Audit window: 12 months ending 1 May 2026. Reviewed quarterly. Detection classes outside this table are not yet ready for public claims.
Where ESA fits best, and where we scope carefully
ESA works best when the motor signal contains stable, repeatable evidence of asset behaviour. Some faults appear directly in current and voltage. Others appear through changes in torque, speed, load, or electromagnetic balance. SAM4 scopes each fleet before rollout so detection confidence is clear before deployment.
Strong fit
Hard-to-access motor-driven assets
- Submerged, enclosed, hazardous-area, remote, or distributed assets
- Pumps, fans, conveyors, mixers, and other motor-driven equipment
- Fleets where asset-mounted sensors are impractical or uneconomic
- Electrical faults visible directly in current and voltage
- Process and load-path faults such as blockage, cavitation, air lock, and mechanical unbalance
- Continuous monitoring from the motor control cabinet
Conditional fit
Good candidates, but scoping matters
- VFD-driven motorsSupported. Drive harmonics can reduce signal-to-noise for some signatures. Drive setup and signal quality are reviewed during scoping.
- GearboxesDetection improves when gear ratio, output shaft speed, and tooth count are known. Without this data, SAM4 may flag abnormal behaviour without naming the exact gear frequency.
- Very low-speed assetsDetection windows are longer and some signatures are weaker. Assets below roughly 300 RPM require review.
- Highly variable processesChanging load, flow, or operating mode can mask weak fault signatures. Baseline quality matters.
- Bearing faultsEvidence strength depends on bearing location, asset type, load path, and operating regime. Motor-side signals are generally stronger than driven-side.
Not supported or poor fit
Where ESA is not the right primary method
- Single-phase motors
- Assets without safe access to the MCC or drive cabinet
- Assets with too little operating time to build a useful baseline
- Failure modes that do not measurably affect current, voltage, torque, speed, load, or electromagnetic balance
- Cases where required asset configuration data is unavailable and fault classification must be precise
Fit is assessed before rollout.
SAM4 reviews asset type, motor configuration, drive setup, operating regime, signal quality, connectivity, and target failure modes before deployment. Where confidence is lower, this is reported upfront.
Why operating context matters as much as the signal
A single measurement tells you what the signal looked like once. Continuous monitoring tells you whether it repeats and how fast it grows. But growth alone is not a fault: the same motor draws different current at different speeds and loads. SAM4 estimates speed and load from the same waveform it analyses, and compares each new spectrum against the asset's own healthy baseline for that operating state. Three things matter together: the trend, the rate of change, and the operating point.
60% load
95% load
Same pump, different load. Different spectrum.
Speed from supply frequency. Load from active power. SAM4 baselines per operating state, so a process change does not look like a fault.
Months to plan
Act now
The trend tells you the urgency.
One measurement says a fault exists. The continuous curve tells you how fast it is developing. Periodic visits capture one condition; continuous monitoring captures the envelope.
Defect indicator over time
A flat line is not recovery.
Mechanical degradation rarely runs in a straight line. The plateau is the defect surface stabilising before it grows again. SAM4 reads non-linear patterns as continued degradation.
Where ESA adds visibility beyond vibration
ESA and vibration are complementary. Vibration is powerful when a sensor can be mounted close to the fault path. ESA adds visibility where the asset is hard to reach, the signal path is electrical, or the fault changes load, torque, speed, or power quality. Many customers use both: vibration for accessible high-criticality machines, ESA for hard-to-reach assets, electrical faults, and fleet-scale coverage.
Vibration
ESA
Measures mechanical acceleration at the asset
Measures current and voltage at the MCC
Strong near the sensor
Strong when faults affect electrical behaviour
Needs physical access
No routine asset access
Best for many mechanical faults
Adds electrical, process, and hard-to-reach coverage
01 / Electrical faults
Electrical faults appear earlier in electrical data
Winding degradation, voltage imbalance, phase imbalance, and rotor bar defects often appear directly in current and voltage. Vibration may only see secondary effects later, if the fault begins to affect torque, heat, or mechanical behaviour.
02 / Hard-to-reach assets
No routine access to the asset
Vibration sensors need a physical mounting point on or near the machine. ESA measures at the motor control cabinet, so submerged pumps, enclosed motors, ATEX-zone assets, and remote equipment can be monitored without routine asset access.
03 / Weak mechanical signal paths
Less dependent on vibration transmission
Vibration depends on mechanical energy travelling from the fault to the sensor. Couplings, housings, fluid, structure, mass, and mounting quality can attenuate or mask the signal. ESA reads the motor’s electrical response to changes in load, torque, speed, and electromagnetic balance.
04 / Fault-origin separation
Voltage helps separate supply issues from asset faults
SAM4 measures current and voltage together. Supply-side issues often appear first in voltage. Downstream mechanical or process issues often appear first in current. Comparing both helps separate power-quality issues from developing asset faults.
Use vibration when
The asset is accessible and mechanical fault transmission to the sensor is strong.
Use ESA when
The asset is hard to reach, electrical faults matter, or fleet-scale cabinet monitoring is more practical.
Use both when
The asset is critical and you want independent evidence from two physical domains.
ESA is not a replacement for vibration.
It extends condition monitoring to faults and assets vibration often misses: electrical issues, inaccessible machines, weak mechanical signal paths, and load or process changes visible through the motor’s electrical behaviour.
Established science. Field-proven performance.
ESA is not a new signal trick. It builds on decades of motor current signature analysis, formalised ESA standards, and peer-reviewed research. SAM4 applies that science to live industrial fleets and validates performance against resolved customer outcomes.
40+ years
of ESA and motor current signature analysis research
7,000+
assets monitored in industrial environments
95.5%
recall on reported core asset classes
2.1%
false-alert rate after expert review
Methodology
How the field metrics are calculated
Detected faultSAM4 raised an alert and maintenance confirmed the fault.
False alertSAM4 raised a customer-facing alert, but maintenance did not confirm the diagnosed fault.
Missed confirmed faultMaintenance confirmed a fault without a prior SAM4 alert.
Data basis: 1,467 scored incidents, 12 months ending 1 May 2026. Reviewed quarterly. Recall = detected faults / confirmed faults. False-alert rate = false alerts / customer-facing alerts.
See full methodology →Validation layers
Where the science comes from
Scientific origin
Oak Ridge National Laboratory
ESA traces back to motor current signature analysis for inaccessible motor-operated valves in nuclear safety systems. The principle remains the same: electromechanical faults can leave measurable signatures in current and voltage.
Standardised method
ISO 20958
ISO 20958 defines condition monitoring and diagnostics using electrical signature analysis on three-phase induction motors. SAM4 builds on that method for online, fleet-scale monitoring.
Industrial validation
ABB drive integration
ABB embedded Samotics ESA into its ACS880 drive portfolio. For compatible drives, SAM4 can analyse current and voltage signals already measured by the drive.
Research base
Peer-reviewed research
ESA methods have been studied in IEEE, EPRI, and condition monitoring literature for decades. SAM4 combines that physics with field data, asset-specific baselines, and expert validation.
Customer evidence
“We invested heavily in condition-based maintenance: vibration, oil analysis, airflow, the lot. But those approaches can still be snapshots. SAM4 added continuous visibility, and that changes how confident you feel about what’s happening.”
Technical FAQ
The model is trained on field-confirmed faults from 7,000+ assets across 80+ customers. Training data spans multiple industries, motor sizes, drive configurations, and operating conditions. Each new validated detection feeds back into training, expanding the fault library. Scale prevents overfitting to any single site or asset type.
For bearing inner race faults, ESA typically provides 6-8 weeks of lead time before mechanical failure. In comparative deployments (e.g., Schiphol Airport), ESA detected faults that vibration monitoring missed entirely. The advantage comes from measuring the electrical transmission path, which carries fault signatures before they produce measurable vibration.
The motor needs to operate above approximately 30% of rated load for ESA to establish a reliable baseline. Below this threshold, fault signatures are too weak relative to electrical noise. SAM4 automatically flags low-load periods and excludes them from analysis.
SAM4 segments data by operating state. Each state gets its own baseline. When equipment cycles between loads, speeds, or idle, the model compares each operating window against the correct baseline. This is critical for assets like borehole pumps or batch-process compressors that run intermittently.
ESA as a method is codified in ISO 20958:2013 (condition monitoring using electrical signature analysis of three-phase induction motors fed from fixed voltage and frequency supplies). The underlying physics has been published in IEEE and EPRI since the 1980s. SAM4's implementation aligns with ISO 20958 for direct-on-line motors. For VFD-driven motors, SAM4 extends beyond the standard's scope using proprietary algorithms trained on real-world industrial data.
See SAM4 ESA with your own data
Request a demo and speak with one of our reliability engineers. We'll show you what ESA detects on assets like yours.