3DSignals Uses Sound Analytics to Revolutionize Machine Maintenance

Industrial IoT’s sensor technology is unlocking solutions to longstanding maintenance challenges. Traditional upkeep has been slow, costly, and heavily reliant on scarce skilled labor.

Israeli start‑up 3DSignals is turning this around by fusing IoT with machine learning to monitor equipment through sound.

I spoke with Amnon Shenfeld, co‑founder and CEO, to learn how the idea sparked.

“It began on a train ride,” Shenfeld recalls. “With a background in deep learning, I was convinced a neural network could automatically recognise and classify machine sounds, and—after expert verification—deliver actionable insights into equipment health.”

He assembled a multidisciplinary team of data scientists, mathematicians and electrical engineers, and visited a local steel plant.

“We asked the operators what failures they most frequently faced,” Shenfeld says. “Saw‑blade breakage tops the list. They report a 50 % efficiency rate—remarkable in steelmaking—but still waste time and resources.”

They had been operating at 20 % uptime before a management change, with blade failures a major pain point. Each replacement could take 20 minutes, and most lines required three changes a day. That’s more than an hour of downtime per machine, plus the risk of damage when blades shatter.

Existing monitoring—current, temperature—proved ineffective. Shenfeld saw an opportunity for 3DSignals.

Disrupting the Preventive Maintenance Cycle

Conventional maintenance demands onsite engineers to inspect machinery on a rigid schedule. The process is labour‑intensive and reactive.

“Historically, technicians relied on acoustics—cars, pumps, etc.—to judge normal operation,” Shenfeld explains. “A seasoned engineer can tell you if a pump is healthy just by listening.”

3DSignals is the first IoT company to leverage sensor‑based sound analytics for a broad range of machines. By learning the expected acoustic signature of each device and then detecting deviations, the system emulates the expert’s intuition with a deep‑learning model.

As a result, maintenance becomes predictive rather than scheduled, allowing engineers to address issues before they lead to downtime.

The platform is cloud‑based and user‑friendly. Sound samples can be shared directly with equipment manufacturers, offering real‑world data that enhances design and troubleshooting.

Sound‑based monitoring aligns naturally with the energy sector, where turbines often run unattended. Routine checks may uncover faults in entirely different equipment—something previously undetectable.

While academia has long studied acoustic analysis for voice recognition, 3DSignals pushes the frontier with automated hearing for industrial assets.

Even Elon Musk once referred to a mysterious sound to investigate the SpaceX Falcon 9 explosion.

When presenting the concept, Shenfeld asks prospective clients, “Can you tell if your machine is functioning correctly just by its sound?” The answer is always “yes.” With applications ranging from mining to agriculture to autonomous vehicles, 3DSignals is poised to reshape maintenance practices across industries.