Multivariate Time-Series Dataset for Simulated Machine Degradation and Prognostics

This dataset provides synthetic multivariate time-series sensor trajectories for simulated machine degradation, remaining useful life prediction, and failure-mode classification.

The official benchmark dataset is:
generated_single_failure_mode_default_dataset/

Additional controlled scenarios are:
generated_single_failure_mode_high_noise_dataset/
generated_multi_failure_mode_default_dataset/

Training files contain full run-to-failure histories.

Test observed files contain truncated histories and should be used as benchmark inputs.

Machine metadata contains validation truth.

Full test truth files contain hidden future trajectories and must not be used as model inputs for benchmark results.

Code, data mirror, and executable examples:

GitHub repository:
https://github.com/cevahiryildirim/mv-ts-mach-deg-dataset

Kaggle data mirror:
https://www.kaggle.com/datasets/cevahiryldrm/mv-ts-mach-deg-dataset

Python quickstart and generator:
https://www.kaggle.com/code/cevahiryldrm/machine-degradation-dataset-python-quickstart

R quickstart and generator:
https://www.kaggle.com/code/cevahiryldrm/machine-degradation-dataset-r-quickstart