by Luis Magadán, José Roldán, Juan Carlos Granda and Francisco José Suárez
Abstract:
Intelligent fault detection and classification is a cornerstone in prognostic and health management of rotating machinery research. Correctly classifying and predicting rotating machinery faults not only increases productivity in industrial plants, but also reduces maintenance costs for companies. The datasets from real facilities needed to train fault classifiers, are often sparse and composed of few samples due to the expense of provoking faults in real scenarios to obtain the data. This paper proposes the use of the Tabular Prior-Data Fitted Network (TabPFN) model for the classification of faults in rotating machinery. TabPFN is a model which has been pre-trained with a large amount of synthetic data containing a large number of causal relationships. This allows the model to perform Bayesian inference on the data used for training. The advantages of this model are its ability to be trained with limited data without generating overfitting problems and its high speed (if a GPU is available). To evaluate its performance compared with traditional algorithms for tabular classification such as XGboost and Random Forest, three public datasets were used. Results show that TabPFN performs better on average than the other algorithms with limited data, so it is suitable to be deployed in real scenarios when the amount of data available from the monitored rotating machinery is limited.
Reference:
Early fault classification in rotating machinery with limited data using TabPFN (Luis Magadán, José Roldán, Juan Carlos Granda and Francisco José Suárez), In IEEE Sensors Journal, volume 23, 2023.
Bibtex Entry:
@article{magadan2023ieeesensors,
author = {Luis Magadán and José Roldán and Juan Carlos Granda and Francisco José Suárez},
title = {Early fault classification in rotating machinery with limited data using {TabPFN}},
volume = {23},
number = {24},
pages = {30960--30970},
issn = {1530-437X},
abstract = {Intelligent fault detection and classification is a cornerstone in prognostic and health management of rotating machinery research. Correctly classifying and predicting rotating machinery faults not only increases productivity in industrial plants, but also reduces maintenance costs for companies. The datasets from real facilities needed to train fault classifiers, are often sparse and composed of few samples due to the expense of provoking faults in real scenarios to obtain the data. This paper proposes the use of the Tabular Prior-Data Fitted Network (TabPFN) model for the classification of faults in rotating machinery. TabPFN is a model which has been pre-trained with a large amount of synthetic data containing a large number of causal relationships. This allows the model to perform Bayesian inference on the data used for training. The advantages of this model are its ability to be trained with limited data without generating overfitting problems and its high speed (if a GPU is available). To evaluate its performance compared with traditional algorithms for tabular classification such as XGboost and Random Forest, three public datasets were used. Results show that TabPFN performs better on average than the other algorithms with limited data, so it is suitable to be deployed in real scenarios when the amount of data available from the monitored rotating machinery is limited.},
author+an = {3=highlight},
date = {2023},
year = {2023},
doi = {10.1109/JSEN.2023.3331100},
journal = {IEEE Sensors Journal},
keywords = {TabPFN, rotating machinery, fault classification, predictive maintenance, IIoT},
shortjournal = {},
jcr = {4.3 -- Q1 [2022]},
file = {revistas/magadan2023ieeesensors.pdf}
}