A Precise, General, Non-Invasive and Automatic Speed Estimation Method for MCSA Diagnosis and Efficiency Estimation of Induction Motors

Abstract

Efficiency estimation and diagnosis via MCSA require precise knowledge of speed. In an industrial environment, speed must be obtained with a non-invasive, automatic and general method. Recent studies have shown that Sensorless Speed Estimation techniques based on detecting Rotational Frequency Sideband Harmonics (RFSHs) or Rotor Slot Harmonics (RSHs) are best suited to these purposes. RFSHs-based methods are easier to apply as they only depend on the number of poles. RSHsbased are much more accurate due to their wider bandwidth. Yet, their use is not trivial as they require to identify the RSHs family, assign to each RSH its order of the current harmonic (ν) and determine the number of rotor slots (R), a rarely known parameter. This paper ends with this trade-off between accuracy and applicability by proposing a novel RSHs-based technique that, for the first time in technical literature, eliminates the need to estimate the number of rotor slots and provides a reliable and automatic procedure to locate the RSHs family and determine their ν indices. Finally, the method is validated under all types of conditions and motor designs, by simulations, lab tests and with 105 industrial motors, highlighting its high accuracy (errors below 0.05 rpm), and applicability.

An Intelligent Hybrid Feature Selection Approach for SCIM Inter-Turn Fault Classification at Minor Load Conditions Using Supervised Learning

“…CBM has impact greatly in world of production and manufacturing; however, the method of collecting useful signals from these systems influences the ease of processing and analyzing these signals for adequate FDI. The non-intrusive method of data collection for induction motor (motor current signature analysis) has recently attracted more attention and patronage due to its ease of data collection, simplicity, precision, and its nature of not disrupting the operation or performance of the system while collecting useful information from the systems [6], [7]. Furthermore, data collection techniques are not the only tools that influence how efficient a diagnostic model becomes; techniques such as signal processing, feature extraction, and feature selection techniques, as well as the diagnostic or classification algorithm that performs the FDI, must be adequate to ensure the model's optimal performance.…”