An Israeli startup has created an AI that uses sound to diagnose mechanical parts.
Surely you have ever heard a strange noise coming from the engine of your car, or from any machine; from the mixer to a hand saw, and even your own computer.
As long as there are moving parts, they will make some noise, but there are many types of noise; if we are used to it, we can instinctively know that something is about to break, even if we do not know exactly what.
Diagnosing faults by listening to the machine
Something similar happened to the founder of the startup 3DSignals, while he was traveling by train and he started making strange noises; Although nobody in the car knew anything about trains, everyone realized that something was wrong.
He couldn’t help but think that, if there was an engineer on the wagon, I could tell you what was going on and if it was serious; In the end it was not, but this experience gave him an idea.
Thus was born a project that apply the latest advances in machine learning to sound; the idea is that an Artificial Intelligence learns the noises that a machine makes when it is failing, and associates them with different types of errors.
The IA that uses the sound is now ready for production
In this way, the system can periodically listen, and send us a notification to the smartphone when we discover something worrying; According to its creators, if it receives the proper training, this AI is capable of identifying specific problems with 98% accuracy.
To obtain the sounds with the best possible quality, the system has a ultrasound microphone; It is capable of capturing sounds up to 100 KHz, making it about five times more sensitive than the human ear. In this way it is capable of capturing sounds that we would normally miss.
Its creators are already in negotiations with manufacturers to implement the system in cars and factories; It can also be used in industrial machinery. For now, the first demonstrations will take place this month.