Topologies classification employed in LED lamps through supervised learning

This work describes the use of artificial neural nets (ANN or RNA) to classify operating patterns of Light Emitting Diode (LED) lamps that use different topologies in their drivers. The parameters used in classification are power measured, luminous flux, power factor, harmonic current, Total Harmonic Distortion (THD), correlated color temperature (CCT or TCC), color reproduction index (CRI or IRC) and luminous efficiency. The research chose 18 single-base LED lamps with an integrated control device, ranging from 8 to 16 W. Among the 18 samples, 6 feature Flyback topology with protection and 12 feature Buck topology, 6 with protection and 6 without protection. The samples were subjected to voltage variations between −20% and + 20% of nominal voltage, totaling 22 voltage levels and 396 measurements. From the measured data, the implementation of a neural network was carried out through the Toolbox Neural Pattern Recognition (NPRTOOL) of MATLAB that encountered the functioning patterns of control devices that employ buck and flyback topologies, classifying them in a correct manner.