Audio analysis for industrial maintenance

A key part of smart manufacturing and a modern factory approach involves real-time monitoring of machinery operating conditions. Follow the tutorial link at the botton of this article to learn how to use the Peltarion Platform to build a model which identifies machinery operating conditions using audio recordings.

Sound of Silence?

02/ What can be achieved?

Using deep learning to detect malfunctioning machinery in real-time will lead to increased productivity and decreased costs.

03/ The problem

Current maintenance schedules often feature blunt techniques such as blind checks or are centered around machine usage alone. Without specific and reliable information on machinery operating conditions, manufacturers tend to be cautious and undergo more maintenance than is necessary, leading to excessive downtime and lost production. Machines that malfunction before the expected lifetime lead to much more severe failure, involving even more downtime and large costs.

04/ Opportunity for deep learning

Consistent real-time information on machinery operating conditions enables vast improvements to traditional maintenance schedules. This can reduce the overall downtime of machinery and allow workers to intervene before total failure, decreasing reparation time and costs.

05/ How is the AI model implemented?

This specific use case features microphones mounted in key parts of each machine. This can, however, easily be combined with other sensory data such as temperature and accelerometer readings (vibration data).

06/ Data requirements

This model uses audio recordings classified either as functioning or malfunctioning machinery. For our model, we converted the audio recordings to spectrograms and applied an image classification model. A spectrogram shows a visual representation of the intensity of different frequencies present in a signal. Click below to compare the audio recording of an industrial solenoid valve to its spectrogram.

Mel-spectrogram of a solenoid valve. Low frequencies at the top and time progress from left to right. The lighter section at the top is the low background noise from the factory and the sharp vertical lines are the distinct clicks from the valves.

07/ Test it yourself

Use the app below to test the functionality of a model we built using the Peltarion Platform to detect whether real industrial machinery is operating under normal or abnormal conditions. Click here to download 10 demo-spectrograms. These spectrograms are a visual representation of audio recordings of valves in a factory environment.

Click the icon below to upload a spectrogram and click the button to classify. Once you have uploaded an image, you can click it again to try another one.

08/ Where to learn more?

A model for this use case has been developed using the Peltarion Platform. The audio recordings are preprocessed to create spectrograms which can be used in a binary image classification model. Follow the machinery operating conditions tutorial to see how this is done.

If you're interested in learning more about AI in manufacturing, have a look at the in-depth article about predictive maintenance or read about how one of our customers uses deep learning for manufacturing to improve paper production.