Smart cities

Direct current arc faults detection

Build your own arc fault detection mechanism with edge AI.

Direct current arc faults detection Direct current arc faults detection Direct current arc faults detection
Direct current arc faults detection Direct current arc faults detection Direct current arc faults detection
Contact us to learn more


Smart cities

NanoEdge AI Studio

Predictive maintenance

Current sensor

Contact us to learn more
Did you know that one out of four fires is triggered by an electrical issue? Most of them are caused by electrical arcs. That is why it is critical to detect them and react as quickly as possible.

As the market evolves, we see new and different types of applications emerging that request more & more innovations to protect against electrical arcs: solar panels, batteries, power tools, e-bikes and so on.

Rule-based algorithms exist today to improve the safety of electrical installations, but their adaptability is limited, and they generate a large number of false positives, thus reducing machine yield. There are also Cloud-based AI solutions, which offer greater accuracy but come with concerns about latency and privacy.

This is where edge AI solutions offer the ideal compromise. Requiring no connectivity or external processing, these solutions enable instant detection and response, while eliminating privacy and security concerns since data is processed locally on devices. Their ability to continuously learn and adapt to different environments also reduces false positive rate and improves efficiency.

In this use case, we will show you how you can simply build your own arc detection mechanism with edge AI and STM32!


Here, we opted for the NanoEdge AI Studio tool due to its user-friendly interface and ease of use. NanoEdge automatically selects and generates the most appropriate model based on the user's data. If we had a pre-trained neural network model, we could have used STM32Cube.AI to compress and optimize it for use in an embedded environment.

Our approach was as follows:
  • We used a custom arc fault circuit interrupter (AFCI) board with an STM32G4 at the heart of it for this demonstration.
  • We first collected the normal operating data from the setup, totaling around 1000 signals. 
  • Then, we gathered arc fault data, also around 1000 signals. 
  • Both sets of signals were imported into a classification project within NanoEdge AI Studio. 
  • The tool then generated the best AI library for this project and we integrated it into the code to monitor the current and trigger an alert if an arc is detected.  


Current Sensor.


2 classes of data Arc fault, No arc
Length data 2048 * 1 axis
Data rate 150 kHz


2 classes (no arc & arc fault):
100% accuracy, 16.7 Kbytes RAM, 0.5 Kbytes Flash
A green point means we are able to correctly predict if the finish will pass a visual inspection or not.
A red point means we were incorrect.

Model created with

NanoEdge AI Studio

Model created with

Compatible with


Compatible with


Model created with NanoEdge AI Studio

A free AutoML software for adding AI to embedded projects, guiding users step by step to easily find the optimal AI model for their requirements.

Model created with NanoEdge AI Studio

Compatible with Any STM32 MCU

The STM32 family of 32-bit microcontrollers based on the Arm Cortex®-M processor is designed to offer new degrees of freedom to MCU users. It offers products combining very high performance, real-time capabilities, digital signal processing, low-power / low-voltage operation, and connectivity, while maintaining full integration and ease of development.

Compatible with Any STM32 MCU

You also might be interested by

Industrial | Transportation

Smart asset tracking

Packages condition classification on sensors.

Industrial | Smart offices

Low-power anomaly detection on a fan

Low-power anomaly detection solution running on a sensor.

Industrial | Smart buildings | Smart homes | Smart offices

Face identification with ID3 Technologies

End-to-end AI solution for face identification running on STM32 microcontrollers.