Name
Title | ||
|---|---|---|
Multilevel DiscreteWavelet Transform and Deep Neural Networks for Predicting Remaining Useful Life of a Machine Asset |
Abstract | ||
|---|---|---|
Estimation of Remaining Useful Life (RUL) of a machine asset using sensor data is a problem that is of prime interest to modern process industries. Accurate estimation of RUL is very critical in prognostics and health management (PHM). Traditional regression-based approaches that utilize both classical machine learning and deep learning architecture do not take into account the features that can be extracted using the temporal characteristics of the time-series sub-sequence pertaining to the sensor data. In this paper, different transformations of these sub-sequences have been used in conjunction with different state-of-art deep learning architectures to significantly improve the accuracy of RUL estimation. The need for such transformations arises when there are no visible trends in the data, leading to the failure of the machine asset. These transformations bring out different critical signatures like sudden bursts of high frequency activity hidden in these sub-sequences. Using extensive experimentation, this paper concludes that if multilevel discrete wavelet transform (DWT) coefficients of the sub-sequences are used as the input feature for the deep learning models instead of the raw sensor values, then a significant improvement in the accuracy of RUL estimation is observed. |
Year | DOI | Venue |
|---|---|---|
2022 | 10.1109/ICPHM53196.2022.9815670 | 2022 IEEE International Conference on Prognostics and Health Management (ICPHM) |
Keywords | DocType | ISBN |
deep learning models,raw sensor values,RUL estimation,deep neural networks,predicting remaining useful life,machine asset,sensor data,prime interest,modern process industries,health management,traditional regression-based approaches,classical machine learning,deep learning architecture,time-series sub-sequence pertaining,different transformations,different critical signatures,multilevel discrete wavelet | Conference | 978-1-6654-6616-5 |
Citations | PageRank | References |
0 | 0.34 | 6 |
Authors | ||
4 | ||
Name | Order | Citations | PageRank |
|---|---|---|---|
| Shivam Bhardwaj | 1 | 0 | 0.34 |
| Prashant Kartikeya | 2 | 0 | 0.34 |
| Ramesh Kumar | 3 | 0 | 0.34 |
| Soudip Roy Chowdhury | 4 | 0 | 0.34 |