A hybrid deep learning model for predicting atmospheric corrosion in steel energy structures under maritime conditions based on time-series data
Atmospheric corrosion of maritime structures remains one of the most challenging issues facing offshore industry. It is well-known that this process significantly reduces the steel strength leading to structural damage with undesirable consequences. Forecasting atmospheric corrosion levels with prec...
Saved in:
| Main Authors: | , , , |
|---|---|
| Format: | Article |
| Language: | English |
| Published: |
Elsevier
2025-03-01
|
| Series: | Results in Engineering |
| Subjects: | |
| Online Access: | http://www.sciencedirect.com/science/article/pii/S2590123025004979 |
| Tags: |
Add Tag
No Tags, Be the first to tag this record!
|
| Summary: | Atmospheric corrosion of maritime structures remains one of the most challenging issues facing offshore industry. It is well-known that this process significantly reduces the steel strength leading to structural damage with undesirable consequences. Forecasting atmospheric corrosion levels with precision is essential for preventing failures, organizing preventive maintenance schedules, and ensuring the durability of structural operations. This study introduces a new hybrid deep learning (HDL) model called Convolutional Gated Recurrent Unit (CGRU) for forecasting atmospheric corrosion in steel structures subjected to maritime conditions using time-series signals based on real experimental setup. By leveraging both the feature extraction strengths of Convolutional layers, which capture spatial hierarchies from input, and the ability of Gated Recurrent Unit (GRU) layers to learn long-term dependencies, the proposed CGRU model can capture both spatial and temporal features of atmospheric corrosion data within time-series signals, resulting in precise predictions. The performance of the proposed CGRU model is compared with that of other state-of-the-art models such as Convolutional Neural Networks (CNN), Gated Recurrent Unit (GRU), Long Short-Term Memory (LSTM), and Deep Neural Network (DNN). The applicability of the proposed model is validated using an experimental time-series corrosion dataset gathered from sensors installed on test site in Gangseo-gu/Busan South Korea. The outcomes of the study will contribute to future monitoring and maintenance concepts ensuring sustainability and safety of maritime structures by providing insights into the practical utilization of deep learning for improving corrosion management in the field of maritime engineering. |
|---|---|
| ISSN: | 2590-1230 |