Performance Benchmarking and Analysis of Prognostic Methods for CMAPSS Datasets

Six years and more than seventy publications later this paper looks back and analyzes the development of prognostic algorithms using C-MAPSS datasets generated and disseminatedby the prognostic center of excellence at NASA Ames Research Center. Among those datasets are five run-to-failure CMAPSS datasets that have been popular due to various characteristicsapplicable to prognostics. The C-MAPSS datasets pose several challenges that are inherent to general prognostics applications. In particular, management of high variability due to sensor noise, effects of operating conditions, and presence of multiple simultaneous fault modes are some factors that have great impact on the generalization capabilities of prognostics algorithms. More than seventy publications have used the C-MAPSS datasets for developing datadriven prognostic algorithms. However, in the absence of performance benchmarking results and due to common misunderstandings in interpreting the relationships between these datasets, it has been difficult for the users to suitably compare their results. In addition to identifying differentiating characteristics in these datasets, this paper also provides performance results for the PHM’08 data challenge wining entries to serve as performance baseline. This paper summarizes various prognostic modeling efforts that used C-MAPSS datasets and provides guidelines and references to further usage of these datasets in a manner that allows clear and consistent comparison between different approaches.