Motivation

Large internal combustion engines form the backbone of many modern energy and transportation systems. In order to meet global challenges such as environmental and climate protection or the conservation of natural resources, the LEC and its partners aim to develop environmentally sound, efficient and robust large engines for sustainable energy generation and transportation applications. Issues with important engine components thwart the achievement of these goals. Yet given the large number of engine components and their interactions, it is often not easy to identify an issue with a particular component. Some types of events only occur for a certain period of time. Nevertheless, cylinder-specific issues may be identified through comparison with signals from the other cylinders of the same engine. To gain a detailed understanding of the root causes of an event and to develop strategies or techniques to avoid specific types of events, engineers rely on either manual or algorithmic issue labeling of available engine data. With the latter, artificial intelligence and in particular machine learning provides powerful tools.

LEC Data Challenge

With the LEC Data Challenge 2023, we would like to encourage students and practitioners of various disciplines who are enthusiastic about data to develop a data-driven solution that identifies cylinder-specific events. A training data set consisting of different cylinder-specific time series data is provided. Consisting of multiple feature signals and an event label, the time series were obtained from large engines in real-world operation. A second data set contains only feature signals, and the goal is to identify the type and time of an issue.

You are free to use any software and programming language to solve the LEC Data Challenge. For this reason, the data sets for training and testing are provided in the versatile CSV format. After completing the LEC Data Challenge, the quality of your solution on the test set will be evaluated in three categories: Overall cylinder event classification, correct label prediction, and temporal precision. You will receive a detailed description of the error/loss measures applied once the LEC Data Challenge has started. The top five teams will be invited to present their approaches to an expert jury during a final (online) event on September 28, 2023.

The LEC Data Challenge 2023 starts on May 23 (earliest data release) and ends August 18, 2023 (submission deadline). Registration and participation are possible throughout the entire period. Data and the detailed task description will be provided for download after successful registration. If you have any questions, please contact us at data.challenge@lec.tugraz.at.

Rewards

The three best solutions selected by an expert jury will receive attractive prize money:

1^st place: EUR 4.000,-

2^nd place: EUR 2.000,-

3^rd place: EUR 1.000,-

We wish you all the best in number crunching!