PHM Data Challenge (

The PHM Data Challenge was part of the PHM08 conference. It was a competition open to all potential conference attendees. The goal was to estimate remaining life of an unspecified component using data-driven techniques. Teams were comprised of one or more researchers. Winners from three categories (determined on the basis of score)d received a $2500 cash prize and were invited to present the successful methods and results in a special session. Results were due June 2, 2008.

Following the competition, the datasets were posted at the data repository of the NASA Ames Prognostics Center of Excellence (link as above). This repository hosts a number of datasets that can be used in a prognostic context.

Original description from the competition:

A data set consisting of multiple multivariate time series is provided. This data set is further divided in to training and testing subset. Each time series is from a different instance of the same complex engineered system (referred to as a “unit”) – e.g., the data might be from a fleet of ships of the same type. Each unit starts with different degrees of initial wear and manufacturing variation which is unknown to the user. This wear and variation is considered normal, i.e., it is not considered a fault condition. There are several operational settings that have a substantial effect on unit performance. These settings are also included in the data. The data is contaminated with sensor noise. The unit is operating normally at the start of each time series, and develops a fault at some point during the series. In the training set, the fault grows in magnitude until system failure. In the test set, the time series ends some time prior to system failure. The objective of the competition is to predict the number of remaining operational cycles before failure in the test set, i.e., the number of operational cycles after the last cycle that the unit will continue to operate.

Algorithms will be scored based on the error of the predictions for the test set. Predictions far from the target are penalized exponentially. The penalty function is asymmetric, with late predictions penalized more heavily than early predictions (i.e., it is better to predict failure too soon than too late). Lower scores are better; a perfect algorithm would score zero. Check the best performing scores here: (

Teams may upload results as often as they like but only the first set of results will be scored from each team each day (defined as 12:00 am to 11:59 pm PDT). Each newly submitted file overrides their previous file. Results may be uploaded until 11:59 pm PDT on 2 June 2008.

Questions may be submitted to Answers will be posted to the FAQ at ( .

Schedule for PHM Data Challenge

2 June 2008 Results due

13 June 2008 Winners announced. Invitation to submit paper

21 July 2008 Papers due

28 July 2008 Reviewers’ comments back to authors

15 August 2008 Final Paper Due


Add New Comment

Popular Resources

Nothing to see here at the moment. Check back later.


Started: Sep 09, 2010

Last Activity: Jul 12, 2017

2 members

What can I do on this project?
  • Any registered users can join
  • Anybody can view this project
  • Any registered users can leave comments
  • Anybody can view comments
New Member

Need help?

Visit our help center