PHM’10 Data Challenge

It is an interesting experience of participating PHM’10 data challenge without any prior background on the machine wear mechanism. Finally, the competition came to an end and I was close to the winning team by less than 5% in score while Mr. Liu won the price for student category. I’d love to see how gtl-phm managed to design the winning prediction algorithm, but at the end of the day, one may find out that the winning strategy could be to exploit the scores from multiple entries prior to the final submission and apply reinforcement learning with the best educated guess of initial wear for each test case. Anyway, I feel hesitating to attend PHM’10 and write a paper about this without knowing the ground truth and what other competitors did especially when the scores between #1 and mine are so close.