Embedding a Neural Classifier to Detect Faults in a Three-Phase Induction Motor

• Renan Gomes Vieira IFCE
• Rebeca Guerreiro C. Cunha IFCE
• Cláudio Marques Sá Medeiros IFCE
• Elias Teodoro Silva IFCE

Resumo

In the literature, a neural classifier design is usually evaluated using only a computer simulator. There is little discussion detailing the rest of the process until the very end step, when the embedded system is implemented. This work aims to fill this gap by analyzing the flow from a classifier, already designed and validated, to its implementation on an embedded platform with resource constraints. The classifier accuracy is evaluated when adopting different strategies to simplify data acquisition process, aiming to reduce resource usage in the target platform. To validate the approach, an implementation is made in a DSP microcontroller and the accuracy results are compared to that obtained by computational simulation.

Palavras-chave: Circuit faults, Induction motors, Biological neural networks, Windings, Frequency conversion, Training, Microcontrollers