MCSA

MCSA for VFD-Driven Motors

Why variable frequency drives complicate motor current signature analysis and what approaches reliably detect rotor bar faults, eccentricity, and stator faults on VFD-driven machines.

Key Constraint

A VFD's output is not a sinusoidal current at one frequency. It is a pulse-width modulated waveform containing energy at the fundamental output frequency plus the switching frequency and its harmonics, sidebands, and intermodulation products. Standard MCSA applied to this signal produces a spectrum full of drive artifacts that can mask or mimic motor fault signatures.

Frequency (Hz) → Amplitude (dB) -20 dB -40 dB -60 dB noise DOL Motor (60 Hz fixed) 57 Hz 60 Hz 63 Hz f1-2sf1 f1 DOL at 60 Hz f1+2sf1 VFD Motor (45 Hz output) 42 Hz 45 Hz 48 Hz fout-2sf fout VFD at 45 Hz fout+2sf sidebands shift with fout VFD-driven motors have a variable fundamental frequency. Fault sidebands shift with the output frequency, requiring analysis referenced to fout not the fixed 60/50 Hz line frequency.

Why VFDs Create MCSA Problems

Standard MCSA assumes the current waveform contains a dominant supply frequency component (50 or 60 Hz) with fault sidebands at predictable offsets. VFDs break this assumption in two distinct ways.

First, the fundamental frequency varies with the drive's output frequency setting, not the utility supply frequency. A motor running at 45 Hz output does not produce fault sidebands at the same locations as a motor running at 60 Hz. The entire fault signature map shifts with the drive command.

Second, the PWM switching frequency, typically 2 to 16 kHz depending on drive model and configuration, generates harmonics, sidebands, and intermodulation products that populate the spectrum in the same frequency ranges where motor fault signatures appear. The result is a spectrum that looks busy regardless of motor condition, making reliable fault identification difficult without additional signal processing.

These two problems are independent. Even a drive running at a fixed output speed still injects switching artifacts. Variable speed adds a third problem: sidebands sweep through frequency as speed changes, making spectrum averaging meaningless unless the analysis is synchronized to the instantaneous drive output frequency.

The Shifting Fundamental

On a VFD-driven motor running at 45 Hz output, rotor bar sidebands appear at 45(1 ± 2s), where s is the slip at that operating point. On a 60 Hz motor with the same load, the same sidebands appear at 60(1 ± 2s). These are different frequencies, and any monitoring system that locks its analysis to 60 Hz line frequency will calculate incorrect fault frequency locations for both cases.

Rotor bar fault sidebands (fixed speed): f_rb = f_out × (1 ± 2s) Where: f_out = VFD output frequency (Hz) s = per-unit slip at operating load Example: f_out = 45 Hz, s = 0.025 Lower sideband: 45 × (1 - 0.05) = 42.75 Hz Upper sideband: 45 × (1 + 0.05) = 47.25 Hz

The practical consequence: if the drive speed is not precisely known at the time of measurement, the expected sideband locations cannot be calculated. At variable speed, the problem compounds further. Sidebands sweep through frequency continuously as speed changes, and a standard FFT spectrum computed over a segment of that data will show those sidebands smeared across a band rather than resolved as distinct peaks.

Any MCSA measurement taken on a VFD-driven motor requires knowing the instantaneous output frequency. Without that, the analyst cannot distinguish a fault sideband from a drive artifact at nearly the same frequency.

PWM Switching Harmonics

PWM switching frequencies generate harmonics in the current spectrum. For a drive with a 4 kHz switching frequency, harmonics appear at 4 kHz, 8 kHz, 12 kHz, and continue at regular intervals. In isolation, these are far above the frequency range of most motor fault signatures and would not cause problems.

The issue is intermodulation. The switching frequency and the output frequency mix in the current spectrum, producing sidebands distributed across lower frequency ranges. For a 4 kHz switching frequency and a 45 Hz output frequency, intermodulation products appear at frequencies like 4000 - (45 × n) and 4000 + (45 × n) for integer values of n. Some of those products land close to where rotor bar or eccentricity sidebands would appear.

The result in a standard MCSA spectrum is a dense forest of peaks across a wide frequency range. Rotor bar sidebands typically sit 20 to 40 dB below the fundamental. Drive intermodulation products at similar amplitudes in the same frequency region make peak identification unreliable without knowing which peaks are drive artifacts and which are motor fault signatures.

Higher switching frequencies push the fundamental harmonic higher but do not eliminate intermodulation products in the low-frequency range. Lower switching frequencies reduce intermodulation product density but increase motor heating. The switching frequency is typically set for drive performance, not for MCSA convenience.

Approaches for VFD-Driven MCSA

Three practical approaches address the VFD-MCSA problem. Each involves trade-offs in capability, cost, and operational disruption.

Approach 1: Bypass the Drive for Testing

If the motor has a bypass contactor, switching to across-the-line operation before taking the MCSA measurement removes all VFD artifacts from the current spectrum. Standard MCSA then applies without modification. Fault sideband locations are calculated using the supply frequency, slip is measured directly from the shaft speed, and the full analytical toolset applies.

The limitations are significant. Bypass operation is not possible during continuous process runs. The motor's load condition at bypass speed (typically full synchronous speed) may differ substantially from its normal VFD operating point, so a fault visible under load at reduced speed may be less apparent at full speed with lighter load. Bypass testing is periodic by nature, not continuous.

For machines where VFD-aware continuous monitoring is not practical, scheduled bypass MCSA at planned outages provides a baseline. The interval between tests should reflect the machine's criticality and the failure modes being monitored.

Approach 2: VFD-Aware MCSA Algorithms

Modern condition monitoring platforms designed for VFD applications apply signal processing specific to drive-connected motors. The core techniques are speed tracking, comb filtering, and demodulation.

Speed tracking uses the fundamental output frequency extracted from the current signal itself as the reference frequency for fault frequency calculations. Rather than assuming 60 Hz, the algorithm identifies the true output frequency in real time and recalculates all fault sideband locations accordingly.

Comb filtering removes the drive switching harmonics and their intermodulation products from the spectrum before fault analysis. The filter is built from knowledge of the switching frequency (either configured or estimated from the spectrum) and removes energy at predictable drive artifact frequencies.

Demodulation extracts the amplitude and phase modulation of the current signal at the output frequency. Motor faults modulate the current in characteristic ways. Demodulation separates the fault-related modulation from drive switching content.

These approaches require knowing or measuring the instantaneous drive output frequency. Some monitoring systems read this directly from the drive via communication interface. Others estimate it from the current spectrum itself. Accuracy of the output frequency measurement directly affects accuracy of the fault frequency calculations.

Approach 3: Flux Monitoring

A search coil placed near the motor frame measures stray magnetic flux rather than supply current. The flux signal contains motor fault signatures but the search coil's inherent frequency response attenuates high-frequency switching transients, reducing drive artifact contamination in the measurement.

Flux monitoring is particularly effective for rotor bar detection on machines where current-based MCSA is unreliable due to VFD interference. The fault signatures appear in the flux spectrum at the same relative frequency offsets as in the current spectrum, but the signal-to-noise ratio at those frequencies is substantially better because drive switching content has been attenuated by the coil's response characteristics.

Installation requires a search coil wound around the motor frame or placed at a fixed location near the motor. This is an additional sensor compared to current-based MCSA, which uses existing current transformers. For machines that are difficult to monitor via current, the additional hardware is often justified.

What the Drive Itself Can Tell You

Modern VFDs log internal data continuously: output current per phase, DC bus voltage, output frequency, fault codes, and often power factor. This data is available via communication interfaces (Modbus, Profibus, EtherNet/IP depending on drive model) without any additional sensors.

Sustained current imbalance between phases reported by the drive indicates either a motor winding problem or a drive output stage problem. Trending phase current imbalance over time identifies progressive deterioration even without detailed spectral analysis. A drive reporting 2% phase imbalance today and 5% imbalance three months later is telling you something is changing.

Drive-reported power factor trends similarly. Degrading insulation or partial winding failures change the motor's reactive current draw. A consistent shift in power factor at the same load and speed condition is worth investigating.

Drive fault history provides a different kind of data: overcurrent trips, ground fault trips, and thermal overload events. A motor approaching failure often produces ground fault currents or intermittent overcurrents before a catastrophic failure. Drive fault logs capture these events even when no dedicated motor monitoring system is present.

Drive data alone cannot replace MCSA for detecting early-stage mechanical faults like developing rotor bar cracks or slight eccentricity. But for machines where continuous detailed MCSA is not practical, drive data trending is a low-cost monitoring layer that costs nothing extra to implement if the drive communication interface is already connected.

Practical Recommendation by Machine Type

Scenario Recommended Approach
Fixed-speed VFD (runs at one speed 90%+ of time) Take measurement at fixed speed; use VFD-aware algorithm or bypass if available. Standard MCSA with corrected reference frequency is viable at stable speed.
Variable speed (speed changes continuously) Flux monitoring or VFD-aware order tracking synchronized to instantaneous speed. Standard MCSA not appropriate; spectrum averaging without speed tracking produces unreliable results.
Critical motor with bypass available Standard MCSA during planned bypass for detailed fault assessment; drive-reported current imbalance trending for continuous monitoring between bypass tests.
Non-critical motor on VFD Drive-reported current imbalance and fault log trending as primary monitoring. MCSA only at planned bypass opportunities or if drive data indicates deterioration.

The choice of approach should be driven by the motor's criticality, the failure modes being monitored, and whether the process permits bypass operation. For most industrial VFD applications, a combination of continuous drive data trending and periodic bypass MCSA provides adequate coverage without requiring specialized VFD-aware monitoring hardware.

For high-criticality VFD-driven motors where early fault detection is operationally important, VFD-aware monitoring platforms with speed-tracked MCSA and comb filtering represent the most capable option. The additional cost relative to standard MCSA reflects the additional signal processing required to produce reliable results from a PWM-contaminated current signal.

Motor Monitoring That Accounts for Drive Output Frequency

Tractian's condition monitoring uses specialized algorithms for VFD-driven motors, tracking fault signatures at the actual drive output frequency rather than assuming fixed line frequency.

See Electrical Monitoring
ElectricalMonitoring.com is an educational resource for industrial electrical and motor monitoring topics. Content is provided for informational purposes only. Always consult qualified electrical engineers and follow applicable safety standards when designing or implementing monitoring systems. ElectricalMonitoring.com is operated by Tractian.