Surface Corrosion Grade Classification using Convolution Neural Network

Abstract

 Abstract: Corrosion is a prevalent issue in the oil and gas industry. Usually, pipelines made of Iron are used for oil and gas transportation. The pipelines are large and distributed over big fields above the ground, underground and even underwater. Corrosion gets developed because of environmental variables such as temperature, humidity and acidic nature of the liquids. There are different techniques for detecting and monitoring corrosion development, both destructive and non-destructive. Visual inspection is a common technique of surface corrosion analysis, but manual inspection is extremely dependent on the inspecting person's abilities and expertise. The findings of the manual inspection are qualitative and may be biased, may result into the accidents because of incorrect analysis. Corrosion must be accurately detected in early phases to prevent unwanted accidents. This paper will present a computer vision-based approach in combination with deep learning for corrosion classification as perISO-8501 standard. The findings of the assessment are unbiased and in a fair acceptable range similar to the outcomes of the visual inspection.

Deep Neural Networks for external corrosion classification in industrial above-ground storage tanks

“…While many authors have dealt with deep learning algorithms to deal with corrosion detection issues [15], [21], [31]- [35], few scientists have focused on classifying corrosion types in different environments [13], [16], [22], [36], [37]. More importantly, no work has been done to classify corrosion types for ASTs, and using EfficientNet architectures.…”

Identifying Paintwork Deterioration for Image-based Monitoring of Distribution Towers

“…The prevention of corrosion using paint coatings is a common practice for physical infrastructure including bridges, marine vessels, pipelines and electricity towers. Thus early detection of failing paintwork is an important inspection requirement with a wide range of applications for which research on automating image analysis has been reported in the literature [3]- [7]. For corrosion assessment, colour and texture features have been used.…”

Brain Region Segmentation using Low MSE based Active Contour Model and Convolutional Neural Network

“…A CNN comprises of various convolutional layers, pooling layers and completely associated layers followed by one order layer [2]. At the point when the size of the image is given as contribution to the CNN include maps are delivered by convolving the image with the channels [3]. Each guide is sub-tested regularly with mean or max pooling layers.…”