How to Detect Linear Motor Winding Faults Early

Overview of Technical Issues:

The diagnostic device and sensing elements insufficiently detect early-stage winding degradation in linear motors—missing incipient insulation breakdown, partial discharge, and localized resistance changes—allowing faults to progress undetected until they cause harmful thermal runaway, complete winding failure, and unplanned motor shutdown; the goal is to achieve early fault detection capability that identifies winding anomalies before they escalate to catastrophic failure.

Solution directions generated for this problem

Problem Direction 1 :

ImproveDiagnostic signal detection sensitivity
VS
ConstraintMeasurement noise interference level

Inspiration 1 : Cross-domain reference

Application Principle: #2 Taking out (Extraction)
Cross-domain applicability Assess applicability
Honeycomb filter defect detecting method and apparatus
Innovative Solution Refine solution

Differential reference channel noise extraction for partial discharge detection

Extract noise via parallel reference path
How to solve :
  • Install a reference sensor adjacent to the diagnostic sensor (within 50mm) that captures identical electromagnetic switching noise and vibration but is electrically isolated from winding faults—use shielded dummy winding segment with intact insulation
  • Implement real-time adaptive subtraction using least-mean-squares algorithm (convergence coefficient μ=0.01, 256-tap FIR filter) that continuously updates noise estimate from reference channel and subtracts it from diagnostic channel at 100 MHz sampling rate
  • Apply coherence gating with threshold 0.85—only subtract noise components showing coherence >0.85 between channels, preserving uncorrelated partial discharge transients below 5pC while rejecting common-mode interference
Expected Effect : SNR recovery to 18-22dB; sub-3pC detection; 0.08% resistance resolution
Risk Control :
  • reference sensor coupling mismatch
  • adaptive filter divergence under transient loads
  • coherence threshold calibration drift

Problem Direction 2 :

ImproveDiagnostic signal detection sensitivity
VS
ConstraintDiagnostic system complexity

Inspiration 1 : Cross-domain reference

Application Principle: #6 Universality (Multi-functionality)
Cross-domain applicability Assess applicability
Bell plate, atomizer-type cleaning device, and associated operating method
Innovative Solution Refine solution

Multi-function current sensor for integrated winding diagnostics

Integrate winding diagnostics into motor drive current sensor
How to solve :
  • Deploy a single high-bandwidth current sensor (≥100 MHz sampling) at motor terminals to simultaneously capture partial discharge transients, resistance drift via impedance spectroscopy, and bearing vibration through current ripple modulation—eliminating separate thermal, electrical, and acoustic sensor arrays
  • Implement multi-domain signal decomposition: apply 20-100 MHz bandpass filtering to extract partial discharge events (≥5pC detection threshold), inject 1-10 kHz AC test signals (10-50 mA amplitude) during PWM off-periods to measure winding resistance changes via synchronous demodulation (0.1% resolution), and analyze 500 Hz-5 kHz current ripple envelope for bearing fault signatures
  • Establish baseline signature library during commissioning—record multi-domain current profiles under controlled conditions (±5% load variation, ±2°C ambient), then deploy threshold-based anomaly detection (resistance drift >0.15%, discharge rate >3 events/hour, vibration amplitude +20%) to trigger 2-4 week advance warnings without continuous complex processing
Expected Effect : Component count -70%, sub-5pC discharge detection, 0.1% resistance resolution, 2-4 week warning
Risk Control :
  • high-frequency noise coupling to current sensor
  • AC injection signal interference with motor control
  • baseline drift due to aging affects threshold accuracy

Problem Direction 3 :

ImproveAnomaly measurement precision
VS
ConstraintDiagnostic system complexity

Inspiration 1 : Cross-domain reference

Application Principle: #26 Copying
Cross-domain applicability Assess applicability
Systems and methods for diagnosing auxiliary equipment associated with an engine
Innovative Solution Refine solution

Physics-based virtual thermal sensor for winding hotspot detection

Virtual thermal sensing via physics model
How to solve :
  • Deploy lumped-parameter thermal network model with 8-node winding representation—each node calculates temperature from measured terminal current, voltage, and ambient temperature using thermal resistance matrix (Rth calibrated ±5% during commissioning)
  • Inject 10mA AC ripple at 1kHz through existing current sensor to measure winding resistance continuously—copper TCR of 0.393%/°C converts 0.04% resistance change to 0.1°C temperature resolution, achieving ±0.3°C virtual precision
  • Validate model weekly via single embedded fiber-optic probe at end-turn hotspot—recalibrate thermal resistance coefficients if deviation exceeds 0.5°C, maintaining long-term accuracy without dense sensor arrays
Expected Effect : ±0.3°C resolution, 2 sensors vs 15-20 physical, complexity -70%
Risk Control :
  • model parameter drift over 6-month operation
  • AC injection coupling noise into drive control
  • thermal resistance calibration accuracy under varying load profiles

Problem Direction 4 :

ImproveEarly fault detection time window
VS
ConstraintMeasurement noise interference level

Inspiration 1 : Cross-domain reference

Application Principle: #2 Taking out (Extraction)
Cross-domain applicability Assess applicability
Methods and systems for data collection, learning, and streaming of machine signals for analytics and maintenance using the industrial internet of things
Innovative Solution Refine solution

Long-term trend extraction from noisy winding measurements for extended early warning

Apply long-term trend extraction to noisy data
How to solve :
  • Implement dual-timescale filtering architecture: fast sampling at 10kHz captures instantaneous partial discharge and resistance, then apply 168-hour moving average to extract slow degradation trends
  • Deploy exponentially-weighted moving average (EWMA) with decay constant α=0.95 on daily aggregated measurements—insulation drift developing over 14-28 days emerges clearly despite instantaneous SNR below 10dB
  • Establish baseline drift detection thresholds: resistance change ≥0.08% over 7 days or partial discharge frequency increase ≥15% over 10 days triggers predictive maintenance alert, achieving 2-4 week advance warning
Expected Effect : Detection window extended to 21±5 days; SNR tolerance reduced to 8dB; false alarm rate <5%
Risk Control :
  • baseline calibration drift over temperature cycles
  • EWMA parameter tuning for different motor types
  • trend extraction latency vs warning time tradeoff

Problem Direction 5 :

ImproveEarly fault detection time window
VS
ConstraintDiagnostic system complexity

Inspiration 1 : Cross-domain reference

Application Principle: #10 Preliminary action
Cross-domain applicability Assess applicability
Intelligent electronic footwear and control logic for automated pedestrian collision avoidance
Innovative Solution Refine solution

Factory baseline fingerprinting for simplified long-term degradation tracking

Establish comprehensive baseline during factory testing to enable simple deviation detection
How to solve :
  • Perform comprehensive multi-modal characterization during factory acceptance testing—record full electrical impedance spectrum (10 kHz–10 MHz), thermal map (±0.3°C resolution via IR camera), and partial discharge signature under controlled conditions, creating a unique motor fingerprint stored in onboard memory
  • Deploy minimal online sensors (single wideband current probe 20–100 MHz, three thermocouples at end turns) that measure only deviation from baseline—system calculates Euclidean distance between current measurements and factory fingerprint weekly, triggering alarm when deviation exceeds 15% threshold corresponding to 2-4 week fault horizon
  • Implement cloud-assisted trend projection—upload weekly deviation metrics (≤5 kB data) to remote server running physics-based insulation aging model (Arrhenius equation with measured activation energy 0.9–1.1 eV for epoxy systems) that extrapolates degradation trajectory, providing failure prediction without onboard computational complexity
Expected Effect : Detection window 2-4 weeks; onboard hardware reduced 70%; system complexity index ≤1.3× baseline
Risk Control :
  • baseline drift over motor lifetime
  • factory test protocol standardization
  • cloud connectivity reliability
Patsnap Eureka Solution