Reconstructing secondary test database from PHM08 challenge data set.

Area of Science:

• Aerospace Engineering
• Data Science
• Machine Learning

Background:

• The PHM08 challenge dataset, originating from NASA's Prognostic Center of Excellence (PCoE) and simulated using C-MAPSS, is crucial for turbofan engine Remaining Useful Life (RUL) prediction.
• The original dataset is divided into training, test, and final test subsets, with a single submission allowed for the final test results.

Purpose of the Study:

• To introduce a reconstructed secondary dataset derived from the original PHM08 turbofan engine data.
• To provide a robust validation mechanism for RUL prediction models by offering ground truth RUL values from reconstructed trajectories.
• To mitigate the ambiguity and risk associated with the single-submission policy of the PHM08 challenge's final test set.

Main Methods:

• Reconstruction of secondary datasets from the training trajectories of the original PHM08 dataset.
• Derivation of datasets containing individual partial trajectories with known Remaining Useful Life (RUL) as ground truth.
• Utilizing these reconstructed datasets for pre-validation of RUL prediction models.

Main Results:

• A reconstructed database is presented, offering known RUL values for partial turbofan engine degradation trajectories.
• This reconstructed dataset enables robust validation of models developed for the PHM08 challenge.
• The data supports the research data article 'A Neural Network Filtering Approach for Similarity-Based Remaining Useful Life Estimations'.

Conclusions:

• The reconstructed dataset provides a reliable method for validating RUL prediction models independent of the PHM08 challenge's submission constraints.
• This approach enhances the trustworthiness of model development and performance evaluation for turbofan engine prognostics.
• The availability of this data facilitates further research and development in predictive maintenance.