Spectrogram Inversion for Reconstruction of Electric Currents at Industrial Frequencies: A Deep Learning Approach

Area of Science:

• Electrical Engineering
• Signal Processing
• Machine Learning

Background:

• Accurate measurement of current intensity and frequency is crucial in various electrical systems.
• Traditional methods may require direct contact or complex sensor setups.
• Magnetic field analysis offers a potential non-invasive alternative.

Purpose of the Study:

• To develop and present a deep learning-based approach for identifying current intensity and frequency.
• To leverage magnetic field data, specifically spectrograms, for electrical parameter estimation.
• To demonstrate a contactless method for current measurement.

Main Methods:

• Utilized magnetic field measurements from a conductor carrying current.
• Employed a magnetic probe to generate time-frequency spectrograms of the magnetic field.
• Applied a convolutional neural network (CNN) model with spectrogram images as input.

Main Results:

• The CNN model successfully estimated current intensity and frequency from spectrogram images.
• The approach demonstrated contactless current estimation capabilities.
• Spectrograms visually represented magnetic field induction values across frequencies over time.

Conclusions:

• Deep learning, specifically CNNs, can effectively reconstruct current intensity and frequency from magnetic field spectrograms.
• Contactless current measurement using magnetic field probes and deep learning is feasible and advantageous.
• This method provides a novel, non-invasive technique for electrical parameter monitoring.