Session: 01-03 Advanced Analytics for Condition Monitoring and Asset Reliability

Paper Number: 174409

174409 - Comprehensive Condition Monitoring Framework for Three-Phase Induction Motors Using Advanced Signal Processing Techniques & Algorithms 

Condition monitoring of three phase induction motor is vital for ensuring operational reliability
and minimizing unplanned downtime. While there are multiple techniques that are currently
employed for motor condition monitoring such as, Vibration analysis, Current signature analysis
, thermal imaging techniques, stray flux analysis, acoustic analysis etc. it is observed that, each
of these methods work better for identifying particular problem at particular condition, like,
vibration analysis could detect bearing faults and rotor imbalances, but it is impacted by
environmental noise and vibration from other sources, While current signature analysis, is used
for rotor failure analysis, it is found that, current signature analysis performance is impacted by
varying load conditions, and low slip rate. The paper proposes a comprehensive motor condition
mapping framework using the most fundamental signals, Motor current, Winding Temperature
and Vibration. Helping us to detect onset of failures such as, Rotor asymmetry, rotor bar failure,
bearing defects and insulation degradation.
The paper proposes use of advance signal processing techniques like - Hilbert-Huang transform
(HHT), Park vector Modulus Transform (PVM), Time Synchronous Averaging (TSA), Discrete
Wavelet transformation (DWT) and zero sequence Voltage analysis (ZSVA), to extract health
indicators and fault features from motor vitals.
The paper illustrates a Real-time data acquisition synchronized across all three modalities,
ensuring time-aligned feature extraction. Each transform technique is optimized for
computational efficiency and embedded deployment, with pre-processed features stored locally
for short-term analysis and periodically transmitted to a remote historian for long-term health
trending
The robustness of the proposed method is evaluated under various motor loading conditions,
operating speeds, and fault intensities. Fault scenarios include artificial introduction of rotor bar
defects, shaft misalignment, bearing inner and outer race degradation, and thermal overloads.
The performance is validated through confusion matrices, receiver operating characteristics
(ROC), and feature importance rankings using advanced algorithms, Results demonstrate that
combining vibration, current, and temperature signals consistently outperform single-modality
approaches, with early fault detection accuracy improvement in most scenarios
 

Presenting Author: Hardik Sharma Baker Hughes

Presenting Author Biography: Hardik Sharma