SNU Prof. Byeng Dong Youn’s Team Wins 3^rd Place

<?xml:namespace prefix = "o" ns = "urn:schemas-microsoft-com:office:office" />

at PHM Asia Pacific 2019 PHM Data Challenge

Professor Byeng Dong Youn of Seoul National University’ Dept. of Mechanical Engineering and his team won 3^rd Place among 1,600 teams in PHM Data Challenge, hosted by Asia Pacific Conference of the Prognostics and Health Management 2019, held in Liaoning Hotel in Beijing, China from July 22^nd to 24^th. PHM Data Challenge is a global competition in which participating teams must analyze data acquired from actual industrial sites and make prognoses about the system, utilizing PHM (Prognostics and Health Management) technology. Prof. Youn’s team have been selected as Winner in multiple Data Challenge competitions, including the one hosted by 2014 IEEE Reliability Society and PHM Data Challenges in the years 2014, 2015, 2017 hosted by PHM Society. The team’s most recent win was 2017 Data Challenge organized by PHMAP. This year, they competed among around 1,600 participating teams from around the world, showcasing their technological skills. Also, the team was the first non-Chinese participant to win a Data Challenge competition hosted by China’s CAICT.

Large rotating machineries such as compressors, steam turbines, and gas turbines are crucial in energy, chemical, power generation and other process industries. This year’s competition presented a problem where the participants had to predict whether if the rotors, main components in such machineries, would fail and, if so, in what order. Rotors have the possibility of loosening and flying off, due to high rotation speed and intense centrifugal force due to high momentum. It presents a high risk in the safety of machinery operation, and could result in high maintenance cost. Therefore, rotor-related problems comprise one of the main issues in the field of PHM.

This year’s competition began with preliminaries, which took place throughout May and June. Finals were carried out for one week in July. During the preliminaries, participants were given a training dataset in which there are 25 cases of normal runs and faults, all ordered, from 5 timepoints in each of the 5 motor modes. Using this dataset, they had to predict and compile a test dataset of 5 modes each with 5 timepoints. The teams submit their results every week, get their scores, and update the algorithm accordingly. Then, at the finals, they are given the final dataset composed from 5 timepoints in 8 motor modes. They have a single chance to predict whether faults occurred and if so in what order. Then the top 8 teams with the highest scores must do a presentation in China, and the final ranking is decided with the added presentation scores.

The ‘tea’ team from Prof. Youn’s lab won the 3^rd Prize. First, the team observed the overall dataset characteristics with speed profile of each data. Then they applied a multidimensional approach in which they extracted time domain factors such as data variation, RMS, peak, mean, RMS ratio and frequency domain factors such as FFT and quefrency. Using this method, they were able to produce perfectly accurate predictions during the preliminaries.

CAICT, organizer of Data Challenge, interviewed the members of team ‘tea’ in celebration of them being the first foreign team to win an award in China. Team ‘tea’ is comprised of 2 PhD students and 1 Master’s student. They explained that their team name came from their wish to win the prize with a relaxed mindset, similar to having a tea-time in the middle of busy student life in graduate school. They added that they would gladly participate in the next competition if there is an opportunity.