42 lines
1.8 KiB
Markdown
42 lines
1.8 KiB
Markdown
---
|
||
layout: single
|
||
title: "Detection and localization of leakages in water networks."
|
||
categories: projects
|
||
excerpt: "We researched the possibilities of leakage detection in real-world water networks in Munich’s suburbs."
|
||
tags: acoustic anomaly-detection
|
||
header:
|
||
teaser: assets/images/projects/pipe_leak.png
|
||
role: Data Scientist, Machine Learning Expert
|
||
skills: Real-world model application
|
||
---
|
||
{: .align-left style="padding:0.1em; width:5em"}
|
||
|
||
Collaborating with Munich City Services ([Stadtwerke München (SWM)](https://www.swm.de/)), our project focused on detecting leaks in water networks. We equipped Munich's suburban infrastructure with contact microphones to capture the sounds of potential leaks.
|
||
|
||
<center>
|
||
<figure class="half" style="max-width: 70%; text-align:center;">
|
||
|
||
<img src="\assets\images\projects\pipe_leak\1fe8265e-ff21-4e9c-8a2a-2ebcaec41728.jpeg" style="margin-bottom: 0em;">
|
||
|
||
<img src="\assets\images\projects\pipe_leak\8d2364f1-1b03-480d-9ed3-09d548f47383.jpeg" style="margin-bottom: 0em;">
|
||
|
||
</figure>
|
||
</center><br>
|
||
|
||
Our study highlighted technical challenges but also provided key insights. By transforming audio into mel spectrograms, we discovered that deep neural networks could identify crucial features more effectively than traditional machine learning methods, leading to further research publications.
|
||
|
||
<center>
|
||
<figure class="half" style="max-width: 70%; text-align:center;">
|
||
|
||
<img src="\assets\images\projects\pipe_leak\5ea8f0ee-6c61-4835-944c-b77683d397ca.jpeg" style="margin-bottom: 0em;">
|
||
|
||
<img src="\assets\images\projects\pipe_leak\cc01cc58-d3f6-4220-b4f4-c7ea26b3a116.jpeg" style="margin-bottom: 0em;">
|
||
|
||
</figure>
|
||
</center><br>
|
||
|
||
This project was active from late 2018 to early 2020.
|
||
|
||
|
||
{% include reference.html %}
|