Skip to content

Engine Fault Detection by Audio Analysis

About


I am currently completing an MCS in Computer Science. My masters dissertation is on Engine Fault Detection using audio analysis. The goal of the project is to find out, to what degree can Internal Combustion Engine health can be determined by using a smartphone microphone.

The idea originated from a tappet sound on my infamous 2002 Audi TT. At the time, I didn't have enough knowledge or experience in the field to identify the sound. However, an expert mechanic could instantly identify the tappet sound and recommend a fix.

At the most basic level, this project could prove useful to create a tool that identifies basic faults for a non-mechanic. If the project is more successful, it may very well be able to identify faults from sound that even an expert mechanic is unable to notice.

Timeline


 

RPM Detection

Completed 12.10.2020

To acquaint myself with audio analysis techniques, my first task is to determine the RPM of the engine from its sound. I opted to perform a Fast Fourier Transform on the audio and find the fundamental frequency of the cylinders firing to determine the RPM.


 

Misfire Detection

Completed 27.10.2020

Misfires occur when the fuel and air mixture in a cylinder doesn't combust. This can be caused by spark plug, injector or coilpack faults. To find a misfire, I used a Fast Fourier Transform and used harmonics of the fundamental frequency of the cylinders and the presence of intermediate frequencies between the harmonic frequencies to determine a misfire.


 

Relative Compression

In progress

My next challenge is to analyse the sound of the starter motor turning an engine over to determine low compression in a cylinder. The idea is that if there is low compression in a cylinder, less load is put on the starter motor allowing it to spin faster with a noticable difference in sound.


Updates


Whitepaper Review: How Shazam Works - Part 2

© 2021 by mcand.ru. All rights reserved.