Corrosion Detection and Prediction for Underwater pipelines using IoT and Machine Learning Techniques

Abstract

Pipelines are commonly utilized to transmit chemical fluids over thousands of kilometres all over the globe. The pipes are designed to withstand a variety of environmental loading conditions, providing safe and durable delivery from the manufacturing location to the coast or distribution station. Leaks in piping systems, on the other side, are among the primary causes of numerous damages for pipeline operators and the surroundings. Pipeline failures may cause significant environmental catastrophes, human deaths, and financial losses. Significant research has been devoted to corrosion and localization using alternative strategies to avoid this threat and preserve an efficient and proper transmission infrastructure. This paper proposed a corrosion detection and prediction system using Internet of Things (IoT) and machine learning techniques. The system collaborates with two different methodologies, such as IoT utilized to collect data from underwater pipelines and various learning algorithms to identify corrosion possibilities. We have used analogue sensors such as thickness, GPS, pH, etc., to capture the current event. Based on pH value, impact of pipe thickness for a specific period has been analysed depending on learning algorithm. The standard defines policy rules and has used a semi-supervised learning algorithm for validation. The Q-learning based classification algorithm generates reward and penalty for each event and, based on that, defines the possibility of corrosion. A variety of extraction of features and selection methods were used during this research using the IoT model. An extensive experiment analysis of the proposed algorithm obtains higher classification and detection accuracy over the traditional machine learning classification algorithms.