Sound patterns as diagnostic evidence
How Sound Analyzer brings recorded vehicle sounds into a technician’s diagnostic workflow.
A project from peoly’s published portfolio.
The challenge: make a sound useful
A vehicle can reveal changes in its condition through sound. Turning that signal into useful diagnostic information requires more than a recording: the technician needs a way to compare it with recognised patterns and understand the result.
From a mobile recording to a learned pattern
peoly developed a machine-learning engine for classifying recorded sounds. The published platform uses a custom densely connected convolutional network and Microsoft Azure services for the audio-analysis workflow.
In the published Škoda Auto use case, technicians use a standard smartphone or tablet to record a vehicle. The application combines sound analysis with vehicle-specific parameters and the car’s usage profile to support assessment of wearing parts and servicing needs.
Supporting the person making the diagnosis
The original story describes sound patterns associated with components including the steering system, air-conditioning compressor and direct-shift gearbox clutches. Capture, classification and a usable mobile interface are brought together in one workflow.
This case shows how audio can become another source of evidence in maintenance. For a different machine or environment, the relevant sound classes, recording conditions and diagnostic workflow need their own assessment.
A similar challenge in your operation?
Tell us about your data, devices and the workflow you want to improve.