Home

Update 8/9/2020: A participating team, IRI, used the online code to reproduce their results, and realized that they had sent us their solutions in the wrong date format. As a result, their scores were not computed correctly using the evaluation system. The BattleDIM Committee confirmed the mistake and decided to allow the re-evaluation of their results considering the correct date format. 

We are pleased to report that the IRI team (Luis, Romero, Joaquim Blesa,  Débora Alves, Gabriela Cembrano, Vicenç Puig, and Eric Duviella) score is €210,772, which puts them in the winning teams! The Committee has decided to jointly award the 3rd place, to both the IRI team and the LeakBusters team.

As a reminder, the Committee had decided to use the economic objective for the purposes of this competition, and also to consider solutions with up to 5% difference (normalized by the highest possible score) as "very close". As a result the first prize (and now the third prize) is jointly awarded to two teams. A more detailed analysis of the results will be provided in the journal paper.
Announcement of the results (3/9/2020)
Congratulations to the winning teams!
First Place (shared)
Tongji: https://youtu.be/dsS3wJP6klc
Under Pressure: https://youtu.be/oPXuyqyboAQ
Third Place (shared)
IRI: https://youtu.be/DnwfCd0RbWI (see update 8/9/2020)
Leakbusters: https://youtu.be/PrHgNnQn-lQ
Honorable Mention
Tsinghua: https://youtu.be/dIAyUhzwAZY
UNIFE: https://youtu.be/Ez0UNwsTs-s
Update 2/9/2020: The papers and presentations have been uploaded on the website. 
Update 26/8/2020: The Battle will be held online via Zoom. Participants can register to the Zoom meeting through this link: https://ucy.zoom.us/meeting/register/tJUkceisrTooGdYW31lWq3x0DKQBdwoLioxD  (This will require your name and email, you can then add a calendar entry to the event. You will get a link to connect to the meeting after this step). Note that the event will also be broadcasted via YouTube through this website.
Update 21/8/2020: The event will take place online on September 3, 2020, at 08:30-10:00 CET (https://time.is/0830_03_September_2020_in_Brussels).  
Update 13/8/2020: All participants should prepare a 5-minute presentation of their paper, in the form of a recorded online presentation. You can use Zoom, Teams, or any other system, which records in an MP4 format (<150Mb, 480px min width, 16:9). We will provide further instructions on how to upload/send us the file. The online event will take place on September 3, 2020 (TBD). We will send an update with instructions within the following days.
Update 16/6/2020: Participant notifications have been sent. Participants should proceed to revise their abstracts if required, and upload them on Zenodo, using the instructions provided at this link.
Update 3/6/2020: We have received solutions from 18 groups! Thank you for your enthusiasm, we understand that this was a very difficult period due to COVID-19. The committee will now review the abstracts and check the result files. We will inform all teams if their submission/abstract is accepted or not, by June 15, 2020.
Update 31/5/2020: Final reminder to submit your results through the "Submission" page, by Monday, 1/6/2020 @ 18:00 (Anywhere on Earth). Best of luck to all participants!
Update 15/4/2020:  The Competition Committee has decided to continue with the competition, despite the change of date of the conference. New updated deadline to submit your reuslts is 1/6/2020. The results will be presented to an online pre-conference workshop during the dates 1-4/9/2020. 
Update 31/3/2020: The coronavirus situation has by now affected everyone around the world, including us, inevitably slowing us down. However, BattLeDIM needs to go on, and its time to release the results template. We look forward to your contributions!
Update 19/2/2020: CCWI/WDSA 2020 has been postponed to 12-15/11/2020, due to the Coronavirus situation in China. The BattLeDIM committee has also decided to extend the competition deadline to May 1st, 2020, to coincide with the submission of the CCWI/WDSA conference abstract.  
Update 4/2/2020: Added deadline date in Rules PDF.
Update 24/1/2020: The 2019 evaluation dataset has been uploaded. 
Update 20/1/2020: Based on the feedback received by participants, we updated the historical 2018 dataset to make it more realistic. The new dataset (Version 1.1) can be found in the "Data" section. The Problem Description has also been modified (version 1.3).

Drinking Water Distribution Networks (DWDN) are susceptible to infrastructure failures, which may lead to water losses. Typically, these water losses are due to background leakages and pipe bursts which may occur anywhere within the distribution network. Background leakages are normally difficult to detect due to their small size, whereas pipe bursts are easier to locate as they are of larger size and may appear on the surface. The early detection and localization of some leakage event is extremely important, as this would reduce the time required for accommodating the event and therefore reducing the risk of further infrastructure degradation, contamination events and consumer complaints. In previous years, a number of methodologies have been proposed to detect and isolate the location of leakage events using various types of sensor measurements. These methods were commonly evaluated on private commercial datasets, and as a result, it is not possible to objectively compare these methods in their ability to detect and isolate leaks. In the past year, a leakage detection dataset has been proposed, LeakDB [1], based on benchmark networks and created using the WNTR tool [2], using pressure-driven demands and realistic leakage modelling [3]. Inspired by the “BATtle of the Attack Detection ALgorithms” (BATADAL), which focused on the detection of cyber-physical attacks [4], our team decided to organize a similar “battle” focusing on leakage events.

The Battle of the Leakage Detection and Isolation Methods (BattLeDIM), organized as part of the 2nd International CCWI/WDSA Joint Conference in Beijing, China (http://www.ccwi-wdsa2020.com/), aims at objectively comparing the performance of methods for the detection and localization of leakage events, relying on SCADA measurements of flow and pressure sensors installed within water distribution networks. Participants may use different types of tools and methods, including (but not limited to) engineering judgement, machine learning, statistical methods, signal processing, and model-based fault diagnosis approaches.

[1] Vrachimis, S. G., Kyriakou, M. S., Eliades, D. G. and Polycarpou, M. M. (2018). LeakDB: A benchmark dataset for leakage diagnosis in water distribution networks. In Proc. of WDSA / CCWI Joint Conference (Vol. 1).

[2] Hart, D., Klise, K.A., Bynum, M.L., Laird, C.D. and Seth, A., (2019). Water Network Tool for Resilience (WNTR) v. 2.0 (No. WNTR). Sandia National Lab (SNL-NM), Albuquerque, NM (United States).

[3] Crowl, D.A., and Louvar, J.F. (2002). Chemical Process Safety: Fundamentals with Applications, 3rd edition. Upper Saddle River, NJ: Prentice Hall, 720p.

[4] R. Taormina et al. (2018). “Battle of the Attack Detection Algorithms: Disclosing Cyber Attacks on Water Distribution Networks,” Journal of Water Resources Planning and Management, vol. 144, no. 8, p. 04018048.