Artificial Intelligence – Sound Analyser
Artificial Intelligence / Sound Analyser
Car Diagnosis
by Sound
The cellphone app helps service technicians to quickly and easily assess what kind of car maintenance is needed
he After Sales department and ŠKODA AUTO DigiLab are testing the new Sound Analyser app. Thanks to artificial intelligence, it helps to rapidly and precisely detect where service maintenance is needed. To this end, it compares data with already available acoustic patterns. In case of inconsistencies, the app estimates what might be the cause and suggests how to correct them. “Sound Analyser clearly shows the options that digitisation is opening up for the ŠKODA brand and in After Sales. In the future, we intend to consistently use AI technologies to provide our customers even more personalised services and give them better care”, says Stanislav Pekař, head of VA – After Sales.
To reliably, unambiguously and quickly assess the condition of worn parts, the app uses various specific parameters to analyse the profile of the respective car. This simplifies the diagnostics for service technicians, as now the only thing they need for their work is a smartphone.
Using this highly sophisticated technology is absolutely intuitive: The smartphone records the car sound by using the app. The next step includes an algorithm comparing the recording with already available patterns providing a specific description of the sound source. The technology recognises 10 patterns with an accuracy of more than 90 percent. They relate to components such as the steering system, the air conditioning compressor and the dual-clutch gearbox (DSG). Currently, the app is “being taught” to recognise other sounds as well.
The project has been trialled in 14 countries since June 2019, including major car markets like Germany, Russia, Austria and France. A total of 245 ŠKODA dealers are participating in this pilot project. First of all, they provide the audio recordings, which the software learning process cannot do without. Gradually introducing technologies for detecting acoustic deviations from the normal status will open up numerous options in sensor-assisted predictive maintenance. Thanks to this and an Internet connection, the car might also automatically detect defects and schedule an appointment at the relevant service workshop if necessary. ED
Technical aspects
The Sound Analyser app is based on a neural network algorithm. First, it converts the audio recording into a spectrogram that visualises the acoustic signals. AI then compares it with the stored values to detect possible deviations. On this basis, the app classifies the need for maintenance or repair according to pre-defined models.