How can you check if an engine is damaged? You can wait until it stops working. Then it is obviously broken. You can take an engine completely apart and examine every single part of it. But then the machine is switched off for a while. You can also listen to the engine while it is working. And this is exactly what the start-up Neuron Soundware does. This company specializes in audio diagnosis of machines. Every machine makes sound while performing its tasks. If there is a mechanical issue, the sound signature of a machine changes. For example, the engine might make extra squeaks, says Pavel Konecny, the company’s CEO and founder. However, Neuron Soundware does much more than what a car mechanic does when they say that something is banging around in the engine. The technology developed by this Prague-based start-up allows you to determine from the engine’s sound if it is working properly. Or if it is close to breaking down. Or even how long it can operate without breaking down. Neuron Soundware uses sound to detect mechanical problems in machines.
How do you do it?
Pavel Konecny, CEO: We use a combination of technologies that use Artificial Intelligence and the Internet of Things to detect changes in sound made by machines in use. We created a system of piezoelectronic microphones mounted on the machine. They sense machine’s movements which we humans can hear as sound. Next, we digitize that signal with our IoT device placed close to the machine. We train the AI to recognize when the machine has any issues and when it works properly. The system learns the machine’s standard movements as well as any anomalies. If the system hears any kind of anomaly, we let the customer about it by email or text message.
At present, we have experience with 16 different types of machines and about 100 IoT devices that have listened to various machines. But it’s not a complete list. We are aiming to add more machines. As each machine type has many different sizes and configurations, we tend to focus on the calibration of algorithms for certain types of machine. If a client comes with a specific machine, we process it on our location, our Acoustic Academy. This is where we collect sounds samples. We record them, test the sensors’ position and after all that, we can add that machine to the list of supported solutions.
There are few companies around the world specializing in acoustic technologies. What is the difference between you and them?
Our competitor from Israel provides a simpler solution than we do. They only focus on production performance and haven’t attempted to address predictive maintenance tasks. When it comes to our rivals from the US, the difference is technical. We are monitoring the machines every second 24 hours a day, seven days a week. Their system can read data once a hour. Also, they rely more on traditional techniques when it comes to accessing machines. Our focus is more on learning from data.
What has been the best moment in the company’s history?
When the CEO of Microsoft Satya Nadella mentioned Neuron Soundware and what we do in his keynote speech that he held in Prague. He could have chose nany other AI startup, but he chose to speak about us. It really was a show of appreciation for us.
What are your plans for a coming year?
We are moving our focus from designing and building a product to focusing on the efficient delivery of our products. Currently, we are considering partners for the manufacture of the device and building up standard delivery procedures. We have to learn how to scale up the standard sales and delivery solutions. We’ve just increased the number of people in our sales team and we are building up all the procedures around sales and support of our customers, partners and distributors.
Where would you like to be in 5 years?
We want to monitor hundred thousands of different machines all over the world. That’s is definitely possible. Our study estimates a possible market of up to 200 million machines and about €65 billion in revenue. So there is really plenty of room where we can scale up.
Are you interested in start-ups? Read more articles on this theme here.