Arc Detection Poster

Detection of Electric Arcs in 42 Volt Automotive Systems Joseph Luis

Markus Zahn

Thomas Keim

Massachusetts Institute of Technology, Laboratory for Electromagnetic and Electronic Systems

Abstract The increasing electrical power demands of automobiles has led to the development of a new 42 V standard. One danger at 42 V is stable electrical arcing that can damage the arcing conductors and cause a fire. This work evaluates using Fourier analysis of the arcing current to detect electric arcs. Fourier analysis is performed on the average of ten arc spectra. This average is compared to the frequency spectra of current during vehicle electric load operation. Results show that there are no easily identifiable features universally present in stable arcs. Also, because the amplitudes of the electric load spectra are of the same order, additional research is needed to find an improved method of identifying arcs.

Frequency Analysis of Arcing Current •DC rise in current is ignored and a 3rd order polynomial fit is subtracted from the arcing current to obtain zero centered arcing transients

Spectra of Automotive Electric Loads •Measured waveforms of current during electrical loads contain large transients comparable to arcing transients •The PWM of loads such as lights are easily recognizable but transients such as door lock operation can be as random as current during arcing.

Generating Arcs • A computer controlled DC motor is used to make intermittent contact between a steel blade and a test wire • The parting of the conductors generates a stable drawn arc • Circuit current is measured by measuring the voltage drop across a shunt resistor

•Example: the periodogram of a door lock current transient has some low frequency structure with slightly more random high frequency content •The amplitude of the door lock spectrum is of the same order of the arcing transients, making amplitude differentiation difficult

•The spectrum of one arc transient shows content at all frequencies •Peaking at some frequencies occurs differently from event to event

• Typical current waveform: (1) a step up in DC current as the battery is shorted through a small resistance (∼20 m? ) (2) a current drop as the conductors separate and arcing begins (3) arcing current continues until conductor gap length is large enough for the arc to extinguish • Result of 42 V arcing: periodic arcing in which the RMS current is lower than the protective fuse rating causes substantial damage to the arcing conductors

•Spectrum varies randomly throughout frequency band, more so at higher frequencies

•The average periodogram over 10 arc events shows broadband energy content •Spectrum falls as 1/f but does not show any other universally identifiable features •Further analysis shows that low frequency components depend on experimental factors such as motor speed

Future Work •Look into other methods of electronic detection such as using bandpass filtering techniques as well as fractal analysis of the random time domain variations of arcing current •Characterize each electrical load’s frequency spectrum and use lookup tables to differentiate load transients from arcing transients

Partial Reference List Wu, Alan, Investigation of Electric Arcs in 42-volt Automotive Systems, MEng Thesis, Massachusetts Institute of Technology, June 2001 Boksiner, J., Silverman, E.J., Arc Detection for Telephone DC Power Distribution Systems, IEEE Telecommunications Energy Conference, 1993

Acknowledgments Professor George Verghese, Alan Wu, Wayne Ryan Project funded by the MIT/Industry Consortium on Advanced Electrical/Electronic Components and Systems