Deep Learning for Underwater Pipeline Corrosion Detection: A Comparative Analysis of CNN Architectures
ABHILASH A
, ADARSH C R , AISHWARYA D , MEGHA , DR. ANITHA T N
Corrosion,VGG16, MobileNet, DenseNet, and ResNet
In this study report, four different Convolutional Neural Network (CNN) algorithms are used to investigate underwater pipeline corrosion detection. As corrosion poses a growing threat to underwater infrastructure, early detection is essential to avert expensive losses and environmental risks. This work intends to investigate CNNs' effectiveness in identifying corrosion in underwater pipes by utilizing their capabilities, which have demonstrated promise in image recognition tasks. The backdrop of underwater pipeline corrosion, the significance of early detection, and CNN algorithms as a potential remedy are all covered in this research. A thorough analysis of the body of research on CNN applications and corrosion detection techniques in related fields is offered. The dataset utilized for testing and training, as well as the particulars of the CNN algorithms used, are described in the methods section. Experimental data and discussions, including comparisons of accuracy, precision, recall, and F1 score, show how well each CNN algorithm performs. By contrasting the suggested CNN-based method of underwater corrosion detection with more recognized methodologies, the study effectively highlights the advantages and disadvantages of the technology. The conclusion includes a synopsis of the key findings of the study, recommendations for other research directions, and implications for underwater pipeline maintenance. Numerous businesses, including oil and gas, telecommunications, and renewable energy, depend heavily on underwater pipelines. But corrosion also endangers a ship's strength, much as weather and time can erode a ship's hull, leading to leaks, damaging the environment, and necessitating costly repairs. Usually, manual examination is required, which is costly, time-consuming, and prone to errors. Consequently, there is an increasing need for trustworthy, economical, and effective methods of automatically identifying corrosion. In recent years, Convolutional Neural Networks (CNNs) have shown to be a very successful technology for image recognition applications. They are perfect for pattern detection in photographs, including underwater camera photos, because of their special capacity to automatically generate hierarchical representations from raw data. The potential of CNNs to address the issues related to underwater pipeline corrosion detection is examined in this study. The study starts with a comprehensive analysis of the body of research on corrosion detection techniques and CNN applications in related domains. Understanding the state-of-the-art now and spotting gaps in the literature that this study seeks to fill are made easier with the help of this review. Building on this understanding, the methodology section explains the four CNN algorithms that were chosen for evaluation, their architectural characteristics, and the dataset that was utilized for training and testing the CNN models. The test results demonstrate how well the CNN-based approach detects damage to underwater pipelines. Each CNN algorithm is assessed using a variety of performance metrics, providing helpful information about its benefits and drawbacks. The benefits of CNNs over antiquated corrosion detection methods are also discussed, emphasizing how quickly and precisely they can assess massive volumes of image data. In summary, our research contributes to ongoing efforts to enhance the maintenance and observation of underwater pipelines. It offers a useful technique for early corrosion identification using CNNs, helping to protect critical infrastructure and the environment. Further research directions could include improving the CNN architecture, looking into new features to improve detection accuracy, and putting the recommended approach into practice and confirming it in real-world situations.
"Deep Learning for Underwater Pipeline Corrosion Detection: A Comparative Analysis of CNN Architectures", IJSDR - International Journal of Scientific Development and Research (www.IJSDR.org), ISSN:2455-2631, Vol.9, Issue 5, page no.632 - 644, May-2024, Available :https://ijsdr.org/papers/IJSDR2405088.pdf
Volume 9
Issue 5,
May-2024
Pages : 632 - 644
Paper Reg. ID: IJSDR_211378
Published Paper Id: IJSDR2405088
Downloads: 000347378
Research Area: Engineering
Country: Bengaluru, Karnataka, India
ISSN: 2455-2631 | IMPACT FACTOR: 9.15 Calculated By Google Scholar | ESTD YEAR: 2016
An International Scholarly Open Access Journal, Peer-Reviewed, Refereed Journal Impact Factor 9.15 Calculate by Google Scholar and Semantic Scholar | AI-Powered Research Tool, Multidisciplinary, Monthly, Multilanguage Journal Indexing in All Major Database & Metadata, Citation Generator
Publisher: IJSDR(IJ Publication) Janvi Wave
