A team of multimedia and software engineers from Kaunus University of Technology (KTU) in Lithuania has developed an artificial intelligence (AI)-based approach for detecting machine failures in manufacturing facilities.

In noisy factory settings, telltale signs of machine failure are not always heard due to noise contamination or interruptions. As such, the KTU team turned to a deep learning approach for monitoring real-life sound data from working industrial machines that does not involve the incorporation of sensors. Instead, machine failure detection is made possible using a microphone pool and a processing device.

Source: Mixabest/CC BY-SA 3.0Source: Mixabest/CC BY-SA 3.0

According to the KTU team, algorithms are trained on a sound dataset called Malfunctioning Industrial Machine Investigation and Inspection (MIMII), which features industrial machine sounds from machines including valves, pumps, fans and slide rails.

"The noise is real manufacturing environment sound that was intentionally blended with pure machine sound at three different SNR — signal-to-noise—levels: 6 dB, 0 dB, and 6 dB. The machine sound was recorded for both normal and abnormal conditions. As a result, we proposed an anomaly detection system for the analysis of real-life industrial machinery failure sounds," explained KTU researcher Rytis Maskeliūnas, co-developer of the invention.

The KTU team suggests that this acoustic anomaly detection approach could help manufacturers avoid unnecessary equipment replacement, reduce maintenance costs, improve work safety and increase equipment availability.

The study, Acoustic Anomaly Detection of Mechanical Failures in Noisy Real-Life Factory Environments, appears in the journal Electronics.

To contact the author of this article, email mdonlon@globalspec.com