skip to Main Content
Airborne Acoustics & Deep Learning: A Combo For Predictive Maintenance

Airborne Acoustics & Deep Learning: A Combo For Predictive Maintenance

Airborne Acoustics & Deep Learning: A Combo For Predictive Maintenance

While using acoustic data in machine maintenance is not new, the use of airborne acoustics is. We speak to Amnon Shenfeld, co-founder and chief executive officer, 3DSignals, on the role this plays in predictive maintenance.

There are various technologies for predictive maintenance such as vibration, temperature, and power consumption analysis, each which shine in their own specific scenarios and applications. However, a relatively new method called airborne acoustics uses data from ultrasonic sensors that stems purely from the sound which is emitted from machines to the air around them. Combine this acoustic monitoring with deep learning technology, and anomalies can potentially be identified and issues predicted before they interrupt production.

Q: Please give a short description of yourself and the company.

Amnon Shenfeld (AS): I am a software engineer by training and have a background in software development, specifically in the embedded signal processing and machine learning space.

The idea for 3DSignals was born on a train ride. I was riding a train in the UK when suddenly there was a loud alarming noise—one that made everyone stop what they were doing and wonder if they should hold onto something.

The noise passed and everything was okay, but it made me think that if an experienced train engineer had been there, he or she would probably be able to tell almost instantly from the noise what the problem was, and most importantly whether it was a sign of danger, and how to fix it after we had reached our destination.

So together with three friends (all experienced engineers), we started to work on this idea around airborne acoustics for predictive maintenance. After developing the technology, we began to test it in a steel production plant that was innovative enough to allow us to experiment around their heavy machinery.

It did not take very long for the system to detect meaningful anomalies and save the factory hundreds of thousands of dollars by preventing unpredicted downtime. Six months after beginning the experiment, we quit our jobs and formed the company 3DSignals with our first design partner.

Q: What does 3DSignals specialise in?

AS: We specialise in the application of sound-based predictive maintenance for industrial and manufacturing equipment. We do this by providing our proprietary hardware and deep learning software system that allows customers to listen to their machines with ultrasonic sensors, anticipate problems with real-time alerts, and reduce unplanned downtime with predictive analytics.

Q: What are some of the key features that distinguish the use of sound in machine predictive maintenance when compared to other technologies?

AS: Condition monitoring via acoustics has three main advantages over other sensing techniques:

Early detection: As changes begin to occur in rotating, mechanical equipment, the subtle nature of ultrasound allows these potential warning signals to be detected early, before actual failure, often before they are detected by vibration, temperature and audible means. In addition, if we zoom out from the individual component out to an entire machine, a single acoustic sensor can detect anomalies in sequence of components operation, where sensors such as vibration are more discrete in nature, isolating each individual component hence missing the “big picture”.

Ease of installation: Acoustic sensors, being external to the equipment, are agnostic and non-destructive. Installation of the equipment takes one day, without taking any machine down, and even without requiring a network access in order to get access remotely. In addition, acoustic sensors are more economical as a single sensor can cover anywhere from a single component to an “orchestra” of machines, compared with multiple or sometimes tens of sensors per piece of equipment.

Sound is intuitive: Plant workers are not data scientists, and often getting their buy-in turns out to be a hurdle, perhaps larger than technology itself. Sound however is intuitive to everyone. Listening to the sound the machine made minutes before it broke, compared to last week’s sound, makes acoustics an effective and friendly root cause analysis method for plant maintenance staff.

Q: While the use of acoustic data is not new in industrial maintenance, what are the modern technologies that 3DSignals are bringing to the table?

AS: Acoustic data is not new in and of itself, but the use of airborne acoustics is new (data that stems purely from the sound which is emitted from machines to the air around them).
The modern technologies we are bringing to the table include ultrasonic sensors developed specifically for harsh industrial environments, as well as innovative deep learning predictive analytics software that learn to recognise and alert on specific acoustic signatures which identify various failure states such as bearing friction, shaft misalignment, blocked filters, lack of lubrication, and more.

Q: What potential changes from the typical machine maintenance process will this entail?

AS: Many existing condition-based maintenance solutions such as vibration analysis via hand-held device require sampling and diagnostics by human technicians going from machine to machine. These contact-based methods can fall victim to producing biased, one-sided results depending on the location of the sensor and experience of the technician and are not constantly monitoring and sending alerts in real time. Other non-hand held sensors with “smart” monitoring capabilities require complex integrations, training and retrofitting of old industrial assets.

Advancements in Industry 4.0 technologies such as the airborne acoustic system powered by deep learning that we developed will increasingly play a larger role in Industrial Internet of Things (IIoT) predictive maintenance. Again, engineers and technicians have always diagnosed machine problems simply by listening to them. However, humans cannot be physically next to every machine at all times during operation and also have a hard time filtering out other noise interference present in harsh industrial environments.

Q: Will the production floor manager be able use this technology separately or in combination with other forms of maintenance for machine tools?

AS: Yes, airborne sound-based predictive maintenance technology can be used both separately and combined with other forms of maintenance—it really depends on the industry and types of machines that are being used. Other methods and technologies for predictive maintenance such as vibration, temperature, and power consumption analysis all excel in their own specific scenarios and applications.

However, in the area of sensor economics, there is only one method that allows for coverage of a wide variety of equipment with a small number of cheap sensors and that method is airborne sound monitoring.

Q: What are some of the insights that have been gleaned from the “hype cycle” for IIoT technologies such as big data?

AS: We actually think that the IIoT hype cycle is still in its early stages in many ways. While there is a lot of excitement about the potential to apply data-driven algorithms to large data streams from industrial assets, IIoT still has a while to go before hitting a mainstream productivity plateau.

One of the main insights we have gleaned from the space so far is that there are still data challenges that need to be overcome to push IIoT predictive maintenance technologies into mainstream adoption and success. The first issue is around data quality the difficulty obtaining high-quality labelled data from industrial machines to begin with.

Secondly, it is then even more challenging to then apply that data to provide human engineers and technicians with relevant and actionable condition-based maintenance insights.

Much of the data quality challenge will be addressed by new deep learning algorithms that mimic the human brain and can be used to build more accurate predictive models. These models will be able to apply insights from previously labelled data to new unlabelled data so that both predictive and prescriptive analysis will become more accurate over time.

APMEN Features

 

Rolling With Additive Manufacturing In The Automotive Sector
3D Printing: From Prototyping To Production
Back To Top