Light Emitting Diodes (LEDs) are the longest lasting source of artificial illumination whose duration can exceed 50.000 continuous working hours. Nevertheless, they show a gradual reduction of the luminous flux due to the increase of the device temperature. In this work, a Machine Learning algorithm will be introduced and discussed, able to predict the junction temperature value of a LED in real-time while connected in the end-user circuit, taking into account current and voltage flowing in the device and, further, the actual model and aging of the LED. The algorithm was implemented on a microcontroller, showing the feasibility of performing edge machine learning on tiny yet powerful devices.
LED junction temperature prediction using machine learning techniques / Merenda, M; Porcaro, C.; Della Corte, F. - (2020), pp. 207-211. (Intervento presentato al convegno 2020 IEEE 20th Mediterranean Electrotechnical Conference ( MELECON) tenutosi a Palermo nel 16-18 giugno 2020) [10.1109/MELECON48756.2020.9140539].
LED junction temperature prediction using machine learning techniques
Merenda M
;Della Corte F
2020-01-01
Abstract
Light Emitting Diodes (LEDs) are the longest lasting source of artificial illumination whose duration can exceed 50.000 continuous working hours. Nevertheless, they show a gradual reduction of the luminous flux due to the increase of the device temperature. In this work, a Machine Learning algorithm will be introduced and discussed, able to predict the junction temperature value of a LED in real-time while connected in the end-user circuit, taking into account current and voltage flowing in the device and, further, the actual model and aging of the LED. The algorithm was implemented on a microcontroller, showing the feasibility of performing edge machine learning on tiny yet powerful devices.File | Dimensione | Formato | |
---|---|---|---|
Merenda_2020_MELECON_LED_Post.pdf
accesso aperto
Tipologia:
Documento in Post-print
Licenza:
Tutti i diritti riservati (All rights reserved)
Dimensione
1.39 MB
Formato
Adobe PDF
|
1.39 MB | Adobe PDF | Visualizza/Apri |
I documenti in IRIS sono protetti da copyright e tutti i diritti sono riservati, salvo diversa indicazione.