Predictive maintenance has become an immensely popular buzzword. Assessing the condition of machines in real time saves time and costs, as the machine can be fixed long before it stops working completely. Neuron Soundware is a Czech startup that proposes an inexpensive sound-based solution. Their product is a small device that, when mounted on a machine, monitors the sounds it makes and warns you about possible malfunctioning.
“Sound is a rich source of data, and also quite universal, which is why mechanics and engineers rely on it so much. But a human cannot listen to 100 airplane or diesel engines for 1000 hours each, and make sense of it all. A machine can do this, and when one engine fails, it can apply that learning to all it has already heard, thus greatly enhancing our ability to detect and prevent future problems,” explained Pavel Konečný, the CEO of Neuron Soundware.
The technology used by Neuron Soundware bases on artificial intelligence and machine learning. The neural network is supplied with hundreds of audio signals (for one client, the team processes about 1.5 TB of training audio files) and learns to distinguish the sounds made by working and malfunctioning machinery. Their solution can be applied to both heavy machinery and sensitive hardware, for example 3D printers, car engines, and air conditioning systems. It already reached an impressive accuracy of 99.5% after just a few seconds of listening to the machinery.
Unlike other similar products, Neuron Soundware can operate offline, on a microcomputer that doesn’t have to connect to the cloud. All data processing takes place in the small device attached to the monitored machine. Thanks to this feature, the hardware is small and inexpensive.
Even though for now the team concentrates on predictive maintenance, their neural networks can also be applied in other fields. For example, their product is capable of reproducing human voices, what makes it possible to replace computer-generated voices of robots or in call centers with more natural ones.
Neuron Soundware was launched in 2016 by two founders: Pavel Konečný, an IT and strategy specialist with a cybernetics background, and Pavel Klinger, a machine learning enthusiast who managed to build his first neural network at the age of fifteen. Konečný, the CEO, has been frequently encouraged to start his own business but decided to wait for about ten years until computers became powerful enough for his needs.
Shortly after the launch, this startup won the Vodafone Idea of the Year 2016 contest against over 170 competing startups from Slovakia and the Czech Republic. They already cooperate with prominent corporate customers such as Siemens (predictive maintenance of wind turbines) or Deutsche Bahn (monitoring of escalators).
Last month, Neuron Soundware received a 600k EUR investment from Prague-based J&T Ventures. This funding will allow them to grow their team, continue with the technology development, as well as reach out to new customers. It will also give the team an opportunity to dive deeper into predictive analysis models.