That is why the MIT has developed a so-called 'smart power outlet'. This is designed to analyse the current consumption of one or more power points and make a distinction between harmless and dangerous electrical arcs.
The device can also be trained to determine what appliance can be connected to a particular power point, such as a fan versus a desktop computer.
The design consists of hardware that processes the electrical signals in real time, and software that analyses this data using a neural network with machine learning algorithms. This has been programmed to determine whether a signal is harmless or not by comparing the measured signal against other ones that the researchers used earlier to train the system. The more information the system receives, the better it can learn the characteristic 'fingerprints' that are used the distinguish good from bad or even distinguish one appliance from another.
Sound cardThe MIT hardware comprises a Raspberry Pi 3, which records the electrical current signals, and an inductive current clamp around the wire to a power point, which detects the current as a changing magnetic field.
Between the current clamp and the RPi is a USB sound card to read the incoming current measurements. Sound cards are, according to MIT, eminently suitable for recording the types of data that is produced by electronic circuits, because they have been designed to capture very small signals at high data speeds.
The sound card also has other advantages, such as a built-in analogue to digital converter with a sampling rate of 48 kHz, which measures signals 48,000 times per second. And an integrated memory buffer, so the device can check the electrical activity in real time.
The smart power point can also make a wireless connection with other devices and therefore take part in a larger IoT network.